KR102647382B1 - Commercial intermediate system - Google Patents

Abstract

In the present invention, a relay service system is disclosed. The relay service system is a relay service system that includes a server that provides a commercial relay service between a large number of buyers and a large number of buyers intensively located in one commercial area, and collects different products to be delivered from different buyers. Sequential clustering is performed for short-distance delivery to be concentrated in one delivery location, and the largest volume product is set as the center of delivery for products subject to delivery or products excluded from the previous stage of clustering. Depending on the distance from the product, sequential clustering is performed to group each product into the same delivery unit, moving forward from near to far, until the volume limit is filled.
According to the relay service system according to an embodiment of the present invention, short-distance delivery is used to deliver a large number of products collected from a large number of buyers intensively located in one commercial area to a single collection location designated by the server operator. In this way, it is possible to provide clustering that supports efficient short-distance delivery that reflects the volume limit suitable for one-time delivery.

Description

Intermediate service system {Commercial intermediate system}

The present invention relates to a relay service system, and more specifically, to a relay service that includes a server that provides a commercial relay service between multiple purchasers and multiple buyers intensively located in one commercial area.

Recently, electronic commerce, in which products or services are traded online, has become active, and accordingly, not only offline buyers but also online buyers operate offline stores in a single commercial area such as a traditional market. There is an increasing need to expand opportunities to contact multiple buyers through There is a growing need to increase buyers' awareness of offline stores, and through this, in addition to revitalizing commercial areas such as markets, shopping experiences in commercial areas such as Namdaemun Market, which is famous at home and abroad, are being commercialized as tourism products. There is a need for research into new relay service systems that can provide this.

One embodiment of the present invention is a short-distance delivery for delivering a large number of products collected from a large number of buyers intensively located in one commercial area to a single collection location designated by the server operator, in one delivery. It includes a relay service system that provides clustering to support efficient short-distance delivery that reflects appropriate volume limitations.

In order to achieve the above and other purposes, the relay service system of the present invention,

A relay service system that includes a server that provides a commercial relay service between multiple purchasers and multiple buyers intensively located within one commercial area,

Perform sequential clustering for short-distance delivery by collecting different delivery target products from different purchasers and concentrating them into one delivery location.

For the forwarding target products or the forwarding target products excluded from the previous stage of clustering, the product with the largest volume is set as the forwarding center, and the volume limit is increased by advancing from the near to the far distance according to the distance from the forwarding center. Sequential clustering can be performed to group each product into the same delivery unit until it is filled.

For example, the server sequentially sets each delivery center and each delivery unit with each delivery center as the center,

For the products subject to delivery, the largest product is set as the primary delivery center, and a primary delivery unit with the primary delivery center as the center is prioritized,

Excluding the above-mentioned primary delivery unit, the product with the largest volume among other delivery target products is set as the secondary delivery center, and a secondary delivery unit is set with the secondary delivery center as the center,

Excluding the 1st delivery unit and the 2nd delivery unit, the product with the largest volume among other delivery target products can be set as the 3rd delivery center, and the 3rd delivery unit can be set with the 3rd delivery center as the center.

For example, the server may request distribution of transportation resources suitable for long-distance transportation of small-volume lightweight products, as the lower-priority delivery unit established in the latter half of sequential clustering includes lightweight products scattered over a wide area. .

For example, the server may not group specific products that exceed a limited distance from the delivery center within the limit volume into the same delivery unit.

For example, the server may group the specific product together into a delivery unit to which other products that are closest to each other belong.

For example, the delivery unit to which the other nearest product belongs may exceed a limit volume while containing the specific product.

For example, the server may request distribution of transportation resources suitable for short-distance transportation of bulky and heavy products for collection units exceeding the limit volume.

For example, the server calculates the difference between the volume of the delivery unit that excludes a specific product exceeding the limit distance and the limit volume,

Depending on the difference between the volume of the forwarding unit and the limit volume, the forwarding unit is released for a forwarding unit whose volume is relatively small compared to the limit volume to the extent that it is judged to be inefficient to bundle the forwarding unit into one forwarding unit. And, the products of the released delivery unit can be grouped into another nearest neighboring delivery unit.

For example, the server groups specific products that exceed the limit distance from the delivery center within the limit volume into the same delivery unit,

For collection units that exceed the above-mentioned distance limit, allocation of transportation resources suitable for long-distance transportation may be requested.

For example, the server releases the delivery unit set in the previous step as a specific product that exceeds the limit distance from the delivery center within the limit volume is captured,

Reclassification can be performed by setting the specific product as the delivery center and resetting the delivery unit with the delivery center as the center.

For example, the server performs preprocessing clustering by sequentially setting a delivery center based on volume and setting a delivery unit with the delivery center as the center, while exceeding a limit distance from each delivery center. Recognize specific products that are not grouped into a delivery unit,

Among the specific products, the product with the largest volume is set as the primary delivery center, and a primary delivery unit is set with the primary delivery center as the center,

Except for the primary delivery unit, the product with the largest volume among other specific products is set as the secondary delivery center, and a secondary delivery unit is set with the secondary delivery center as the center,

If there are no other specific products other than the 2nd delivery unit, the product with the largest volume among the other products excluding the 2nd delivery unit is the 3rd delivery center, and the 3rd delivery center is the 3rd delivery center. You can set the delivery unit.

For example, the server performs pre-processing clustering by sequentially setting a transmission center based on volume and setting a transmission unit with the transmission center as the center, but transmits the transmission beyond the distance limit from each transmission center. Preprocessing clustering can be performed until no other products exist except for specific products that are not grouped into a unit.

According to the relay service system according to an embodiment of the present invention, short-distance delivery is used to deliver a large number of products collected from a large number of buyers intensively located in one commercial area to a single collection location designated by the server operator. In this way, it is possible to provide clustering that supports efficient short-distance delivery that reflects the volume limit suitable for one-time delivery.

FIG. 1 shows a diagram illustrating short-distance delivery and long-distance delivery in clustering for short-distance delivery according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating clustered delivery units in short-distance delivery according to an embodiment of the present invention.
FIG. 3A is a diagram illustrating clustering based on location information and volume information of products corresponding to the location of the purchaser, in one embodiment of the present invention.
In Figure 3b, in a comparative example in contrast to the present invention, a diagram showing the problem of clustering with the local maximum as the center of delivery from the location information and volume information of the delivery target product corresponding to the location of the purchaser is shown. .
In Figure 4, in one embodiment of the present invention, a diagram for explaining clustering for short-distance delivery based on location information and volume information of the delivery target product corresponding to the location of the purchaser is shown.
In Figure 5a, in one embodiment of the present invention, the primary delivery center and primary delivery unit are set based on the location information and volume information of the delivery target product corresponding to the location of the purchaser, and secondary delivery is sequentially performed. A diagram showing an example of clustering that sets the center and secondary transmission units is shown.
In Figure 5b, as shown in Figure 5a, in clustering where the forwarding center and forwarding unit are sequentially set, a forwarding method is used to prevent lightweight products distributed over a wide area from being set as a single forwarding unit in sequentially subordinated clustering. A drawing to explain the reclassification of units is shown.
FIG. 6 is a diagram illustrating a first selection path in which product selection is initiated from information on the product itself, regardless of the location of the purchaser.
FIG. 7 is a diagram illustrating a second selection path in which product selection is initiated from the purchaser's location information.
FIG. 8 is a diagram schematically showing a server in which the product list (link information) provided in the second selection path shown in FIG. 7 is stored in the form of a web document for each purchaser.
FIG. 9 shows a diagram illustrating another second selection path in which product selection begins from the purchaser location for each theme product.
FIG. 10 shows a diagram showing an embodiment of a web implemented for relaying commercial transactions for products set to be private.
FIG. 11 shows a diagram illustrating the one-to-one relationship between a purchaser and a purchaser for disclosing a product set as private.
12A to 12C are diagrams illustrating different classes that can be classified according to external characteristics in accessories such as earrings.
Figure 13 shows a diagram to explain an artificial intelligence neural network model that implements a regression algorithm that analyzes the correlation between multiple attributes and class labels as characteristics of the appearance of a private product.
Figure 14 shows a diagram to explain a convolutional neural network model that predicts the probability that a non-public product belongs to a non-public product using the product image itself as input.
15A to 15C are examples of filters included in the convolutional neural network model, and a high value is applied along the appearance features of each class to generate a feature map that can extract the appearance features of each class belonging to the same product category. There are diagrams showing exemplary filters with weights of .
Figure 16 shows a diagram showing the extraction of index candidate products of the first and second groups according to rapid changes in buyer preferences.
Figure 17 is a diagram showing a series of processes of a server that outputs a trend indicator product image by inputting a group of indicator candidate product images.
Figure 18 illustrates photos showing various children's clothing as an embodiment of fashion items suitable for selection of trend indicator products.
Figures 19A to 19C show diagrams to explain points accumulated, deducted, and managed in a one-to-one relationship between a purchaser and a purchaser.
Figure 20 shows a diagram to explain the structure of point data stored in the server and the registration of points related to offline purchasing activities.
Figures 21a and 21b show diagrams showing an example in which point details of the purchaser and buyer are displayed on the web of the purchaser and buyer according to the account type logged in to the server.

Hereinafter, with reference to the attached drawings, a relay service system according to a preferred embodiment of the present invention will be described. Clustering for short-distance delivery according to various aspects of the relay service system of the present invention, based on the location of the purchaser. We will sequentially explain product selection, privacy settings to prevent imitation of small fashion items, detection of trend changes due to rapid changes in buyer preferences, and points accumulated, deducted, and managed in the one-to-one relationship between purchaser and buyer.

<Short distance home delivery>

First, referring to FIGS. 1 to 5B, clustering for short-distance transmission will be described.

FIG. 1 shows a diagram illustrating short-distance delivery and long-distance delivery in clustering for short-distance delivery according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating a forwarding unit (U) clustered in short-distance forwarding according to an embodiment of the present invention.

FIG. 3A is a diagram illustrating clustering based on location information and volume information of products corresponding to the location of the purchaser, in one embodiment of the present invention.

In Figure 3b, in a comparative example in contrast to the present invention, a diagram showing the problem of clustering with the local maximum as the delivery center (C) from the location information and volume information of the delivery target product corresponding to the location of the purchaser. It is shown.

In Figure 4, in one embodiment of the present invention, a diagram for explaining clustering for short-distance delivery based on location information and volume information of the delivery target product corresponding to the location of the purchaser is shown.

In Figure 5a, in one embodiment of the present invention, the primary delivery center (C) and the primary delivery unit (U) are set based on the location information and volume information of the delivery target product corresponding to the location of the purchaser. , A diagram showing an example of clustering that sequentially sets the secondary transmission center (C) and the secondary transmission unit (U) is shown.

In Figure 5b, as shown in Figure 5a, in clustering that sequentially sets the forwarding center (C) and the forwarding unit (U), lightweight products distributed over a wide area in sequentially lower priority clustering are divided into one forwarding unit (U). ) A diagram is shown to explain the reclassification of the delivery unit (U) to prevent it from being set.

Referring to Figures 1 and 2, clustering for short-distance mailing according to an embodiment of the present invention may be performed by a server that provides a commercial transaction brokerage service between a buyer and a purchaser, and the server provides a service to a plurality of purchasers. Short-distance delivery that collects location information and volume information of products from multiple buyers and clusters products from different buyers into delivery units (U) with a cumulative volume suitable for one-time delivery based on short distance. Clustering can be provided for.

In one embodiment of the present invention, the delivery unit (U) may mean a group of products to be delivered located in a short distance based on the delivery center (C), and a group of products classified into these delivery units (U) Products subject to postage may have a cumulative volume suitable for one-time delivery. Here, the cumulative volume may mean the total volume of a group of delivery target products classified into a delivery unit (U).

In one embodiment of the present invention, the forwarding unit (U) may include a group of forwarding target products arranged in a short distance based on the distance from the forwarding center (C), for example, the forwarding center (C) This may mean a product with a relatively large volume among a plurality of products subject to delivery, and in one embodiment of the present invention, the delivery unit (U) is referred to as the delivery center (C), that is, a product with a relatively large volume is Classifying the delivery unit (U) by centering on products with relatively large volumes means grouping large and small products together into a delivery unit (U) so that they can be shipped once. It is designed to increase the efficiency of mailing. For example, one embodiment of the present invention relates to a short-distance delivery system in which a plurality of purchasers providing different products to be delivered are intensively located in one commercial area, which is recognized as one commercial area such as Namdaemun Market. Based on the region where possible, different products subject to delivery are collected from a large number of buyers scattered in Namdaemun Market, gathered into one delivery location, and then the products subject to delivery are delivered to a large number of buyers located far away from the delivery location. In this sense, in one embodiment of the present invention, short-distance delivery may correspond to the previous stage of long-distance delivery, and after collecting different delivery target products from different purchasers at one delivery location, By initiating delivery to distant buyers, delivery efficiency can be increased. For example, as a deliverer who has been requested by a server operator to deliver from one delivery address to various distant delivery destinations, the Since delivery of the product can be initiated, delivery costs can be reduced, which can ultimately lead to a reduction in purchase costs.

As such, one embodiment of the present invention is a step before long-distance delivery and provides clustering for short-distance delivery, so one delivery unit (U) is not relatively large and, for example, a small transportation vehicle such as a bike The number of delivery units (U) per delivery may be limited, and since a separate small transportation means is used, there may be restrictions based on volume rather than weight. For example, in one embodiment of the present invention, when using a small transportation means such as a bike, the total amount of volume that can be bundled in one delivery unit (U) can be set to approximately 60L, and can be set to a maximum of 90L. You can. The drawings attached to this specification show exemplary transmission unit (U) clustering when the total volume that can be bundled into one transmission unit (U) is set to 60L.

In one embodiment of the present invention, the bulky goods are set as the delivery center (C), and the goods to be delivered located at a short distance from the delivery center (C) are classified into one delivery unit (U), so that relatively bulky goods are classified into one delivery unit (U). By mixing and small-volume products and grouping them into one delivery unit (U), the total volume (60L) limited to one delivery unit (U) can be effectively utilized. For example, by inserting and filling small-volume products between relatively large-volume products, the volume that can be transported in one delivery unit (U) can be efficiently utilized, and in one embodiment of the present invention, the relative By bundling together a large-volume delivery center (C) and relatively small-volume delivery target products located at a short distance from the delivery center (C), the efficiency of collection or transportation can be improved.

In one embodiment of the present invention, the location information of the purchaser and the volume information of the product to be delivered can be obtained to classify the delivery unit (U), for example, through a server that relays commercial transactions between the purchaser and the buyer. Clustering to classify the collection unit (U) may be performed. Here, the location information of the purchaser may correspond to the location information of the product to be delivered, and in one embodiment of the present invention, the location information of the purchaser or the location information of the product to be delivered may include GPS coordinate information of the purchaser. , may include two-dimensional GPS coordinate information.

Referring to FIGS. 3A to 5B, the drawings attached to this specification display the volume of the product to be delivered on a mesh at regular intervals indicating two-dimensional coordinates. In one embodiment of the present invention, it is possible to obtain location information of a delivery target product expressed in two-dimensional coordinates, where the two-dimensional coordinates may mean the location of each building where a plurality of purchasers are intensively located. For example, location information displayed on one GPS coordinate may refer to the location of a building where multiple purchasers are located together, rather than location information of each purchaser. Accordingly, in the present invention, the location information of the product subject to delivery may mean the location information of the buyer who has registered the product subject to delivery, or the location of the building where the buyer is located. In one embodiment of the present invention, the server may store the location of the building where the purchaser is located as location information of the buyer who registered each product to be delivered, and accordingly, a plurality of products to be delivered are stored in one location, So, it can be matched to the location of one building or to GPS coordinates representing the location of one building, and at this time, the server adds up the total volume of multiple mailing target products matched on the same location (building location), and adds up. By treating the volume as a single delivery target product, clustering (two-dimensional clustering) of the delivery unit (U), which will be described later, can be performed. However, in various embodiments of the present invention, the server treats multiple mailing target products matched on the same location (building location) as a single mailing target product with volume added, which means that the volume of the single mailing target product combined by volume Under the premise that it does not exceed the total volume limited to one delivery unit (U) (for example, 60L), it can be treated as a single delivery target product with volume added. In contrast, the volume of the combined single delivery target product is If the total volume (60L) limited to one delivery unit (U) is exceeded, a single delivery target product can form at least two delivery units (U), so the total volume limited to one delivery unit (U) ( Up to 60L) are grouped into one delivery unit (U) or two or more delivery units (U) (one-dimensional clustering based on level coordinates described later), and a number of delivery target products that do not exceed the remaining volume total are grouped into a single delivery target product. By handling, clustering (two-dimensional clustering) of the collection unit (U), which will be described later, can be performed.

In various embodiments of the present invention, clustering (two-dimensional clustering) of collection units (U) based on two-dimensional GPS coordinates as described later, and clustering exceeding the limited volume (60L) matched to one location (building location) For multiple products to be delivered, level coordinates are introduced, and products to be delivered nearby are grouped into one delivery unit (U) up to a limited volume (60L) according to one-dimensional level coordinates (one-dimensional clustering). Except for the forwarding unit (U), the remaining forwarding target products can be treated as a single forwarding target product located at the corresponding two-dimensional coordinates, and two-dimensional clustering described later can be performed.

In one embodiment of the present invention, the location information of multiple purchasers belonging to one commercial area can be stored on a server, purchase information from the purchaser is obtained, and the purchase information recorded in the purchase information is obtained along with the location information of the purchaser. Information on products, that is, products subject to mail delivery, can be matched with each other. For example, the server can determine the location information (location information of the purchaser) and volume information of the product to be mailed from each purchase information, for example, by multiplying the quantity information with the individual volume of the product to be purchased. You can calculate the volume of the product to be delivered from each purchaser.

The server sets the delivery center (C) using the location information and volume information of the delivery target product at the user's request or before the pre-designated time for delivery to start, and sets the delivery center (C) The convergence unit (U) can be classified based on the near distance from the convergence unit (U), and the distance ( By increasing L), clustering can be performed to classify adjacent forwarding target products into the same forwarding unit (U).

The server can set the delivery center (C) using the location information and volume information of a plurality of delivery target products targeting one commercial area where short-distance delivery is performed. In one embodiment of the present invention, a plurality of delivery centers (C) are used. Depending on the volume size relationship of the delivery target product, a delivery center (C) and a delivery unit (U) with each delivery center (C) can be set sequentially.

Referring to FIG. 3A, in clustering for short-distance forwarding according to an embodiment of the present invention, each forwarding center (C) and a forwarding unit (U) with each forwarding center (C) as the center are sequentially set. Through this, each transmission center (C) and each transmission unit (U) with each transmission center (C) as the center can be set sequentially, for example, the primary transmission center (C1) and the primary transmission center (C1). The primary delivery unit (U1) with the center (C1) as the center is set first, and among the other delivery target products excluding the primary delivery unit (U1), the secondary delivery center (C2) and the secondary delivery center (C2) are set first. ) is set as the center, and then, excluding the first delivery unit (U1) and the second delivery unit (U2), the third delivery center (C, Mido) is selected among the other delivery target products. A tertiary delivery unit (U) can be established with the city) and the tertiary delivery center (C, not shown) as the center. In this way, in one embodiment of the present invention, after setting a plurality of delivery centers (C) at once for one commerce zone, a delivery unit (U) is set with each delivery center (C) as the center. Rather, sequentially setting the delivery center (C) and the delivery unit (U) with the corresponding delivery center (C) as the center for the remaining delivery target products excluding the delivery unit (U) set in the previous step. By setting the convergence center (C) and convergence unit (U), the computational burden and computation time required for clustering can be reduced. Referring to FIG. 3b, in a comparative example in contrast to the present invention, a plurality of delivery centers (C) are set at once for one commerce area where short-distance delivery is performed, and each delivery center (C) is set as a center. Units (U) can be classified, but the locations of multiple delivery target products based on purchase information up to each delivery start time may be randomly distributed, so the multiple delivery target products according to two-dimensional locations can be classified. Using the two-dimensional gradient formed by the change in volume (for example, the derivative of the two-dimensional gradient), calculate the local maximum formed by the volume of multiple delivery target products and deliver each local maximum. When using the method of calculating by center (C), each transmission center (C) is set adjacent to each other, and there is often an excess or lack of volume in grouping each transmission center (C) into a transmission unit (U) as the center. Conversely, if each transmission center (C) is set too far away and each transmission center (C) is grouped into a transmission unit (U) with the center as the center, there is a problem between neighboring transmission units (U). There may be a high probability that inefficient clustering will be performed, such as where there may be a large number of forwarding target products that do not belong to any forwarding unit (U).

As shown in FIG. 3A, in one embodiment of the present invention, for a plurality of products to be forwarded, the product with the largest volume is set as the primary forwarding center (C1) according to the size relationship between volumes, and the first After setting the primary delivery unit (U1) with the delivery center (C1) as the center, set the next bulky delivery target product as the secondary delivery target, and set the second delivery unit (U1) with the secondary delivery center (C2) as the center. By setting the secondary convergence unit (U2), clustering can be performed by sequentially setting the convergence center (C) and convergence unit (U) according to the size relationship of the volume.

Referring to FIG. 4, in one embodiment of the present invention, each delivery center (C) selects the product with the largest volume as the delivery center (C) for the remaining delivery target products excluding the delivery unit (U) set in the previous step. ), with each transmission center (C) as the center, and based on the distance (L) from each transmission center (C), the limit of the transmission unit (U) set in advance from near to far. One forwarding unit (U) can be set by repeating clustering to group adjacent forwarding target products into the same forwarding unit (U) until the volume (60L) is filled.

Referring to FIG. 5A, in one embodiment of the present invention, clustering is performed by setting the product with the largest volume as the delivery center (C) among other products excluding the delivery unit (U) set in the previous step according to the size relationship of volume. Therefore, in the later stages of clustering, products of relatively small volume distributed over a gradually wide area are grouped into one delivery unit (U), until all collection and delivery products are classified into each delivery unit (U). Clustering may be performed. For example, in one embodiment of the present invention, the priority delivery unit (U) established throughout the clustering may consider the transportation of products placed at a short distance, focusing on relatively bulky products, and this priority delivery unit For (U), the server may request the server operator to allocate transportation resources suitable for short-distance transportation of bulky and heavy goods. Meanwhile, in the lower-priority delivery unit (U) set in the latter half of clustering, transportation of products placed at a distance, focusing on relatively small and lightweight products, can be considered. For this lower-priority delivery unit (U), the server The server operator can be requested to allocate transportation resources suitable for long-distance transportation of small-sized, lightweight goods.

Referring to FIG. 5A, in one embodiment of the present invention, clustering of products to be delivered is performed for one commerce zone including a plurality of buyers concentrated in a short distance, and is an embodiment of the present invention. As in the form, even if small-volume lightweight products scattered over a relatively large area are grouped into one forwarding unit (U) in the latter half of clustering, the efficiency of forwarding may not be greatly affected. For example, for a low-priority forwarding unit (U) set in the latter half of clustering, the server may request the server operator to allocate transportation resources that can quickly transport lightweight goods to the forwarding location.

In one embodiment of the present invention, the delivery center (C) is set in consideration of the size relationship of the volume of the delivery target product, and clustering based on the short distance from the delivery center (C) can be performed, as shown in Figure 4. , In various embodiments of the present invention, in order to prevent the forwarding unit (U) of the forwarding target product from gradually expanding into a wide area as clustering progresses, the forwarding center (C) is set based on the volume of the forwarding target product. , If the delivery unit (U) exceeds the preset limit distance (Lmax), the limit distance is set by setting the delivery unit (U), excluding products whose limit distance (Lmax) is exceeded from the delivery center (C). It is possible to avoid products exceeding 100% being grouped into the same delivery unit (U). For example, in the second half of clustering, the product with the largest volume among the products to be forwarded, excluding the forwarding unit (U) set in the previous step, is set as the forwarding center (C), and is advanced along the distance direction from the set forwarding center (C). While grouping adjacent delivery target products into one delivery unit (U), this clustering is carried out up to the limit volume (60L) of the preset delivery unit (U), and products placed at a distance from the delivery center (C) are gradually clustered. It can be grouped into one delivery unit (U). At this time, the server can check the volume information and location information of each product by grouping each product into one delivery unit (U) according to the distance (L) from the delivery center (C), and It is determined whether the volume information and location information meet the conditions of the delivery unit (U), and depending on the judgment result, each product can be grouped into a delivery unit (U), and the corresponding delivery unit (U) can be expanded. can be stopped, clustering to the corresponding transmission unit (U) can be stopped, and the corresponding transmission unit (U) can be confirmed. For example, in various embodiments of the present invention, a product (specific product) exceeding the distance limit (Lmax) from the delivery center (C) may be set as the corresponding delivery unit (U2) (see FIG. 5A), Alternatively, products exceeding the distance limit (Lmax) may be excluded from the corresponding delivery unit (U) (see Figure 4). In this case, products exceeding the distance limit (Lmax) may be excluded from the corresponding delivery unit (U). It can be classified into another delivery unit (U1) to which the product exceeding the distance limit (Lmax) and the other closest neighboring product belong (see Figure 5a), and to another delivery unit (U1) to which the product exceeding the distance limit (Lmax) and the other closest neighboring product belong ( U1) may exceed the volume limit (60L) while additionally containing products exceeding the distance limit (Lmax).

In addition, for the delivery unit (U) in which a product (specific product) exceeding the distance limit is captured, for example, each product is transferred to the delivery unit as it advances in the direction away from the delivery center (C) until the limit volume is filled. While performing clustering into (U), for the delivery unit (U) in which products whose distance from the delivery center (C) exceeds the distance limit (Lmax) are captured, the corresponding product is assigned to the corresponding delivery unit (U). Regardless of the decision on whether to classify it as a collection unit (U), the difference between the volume currently bundled into a collection unit (U) and the limit volume (60L) is such that it is judged to be inefficient to group it into a single collection unit (U). For the delivery unit (U) whose volume is relatively small compared to the volume (60L), the delivery unit (U) is released, and the delivery unit (U) is released to the delivery unit (U) to which each product classified by the corresponding delivery unit (U) and other nearest neighboring products belong. It can be classified into units (U).

In one embodiment of the present invention, for a delivery unit (U) in which products exceeding the distance limit (Lmax) from the delivery center (C) are captured, the volume bound to the delivery unit (U) to date is the limit volume (60L) ), if it falls under half (1/2) of the number, it is judged to be inefficient to bundle into one delivery unit (U), so the delivery unit (U) is released, and each product classified into the corresponding delivery unit (U) It can be classified into the delivery unit (U) to which the other nearest neighboring products belong.

In one embodiment of the present invention, the following conditions can be set in advance as the conditions for the transmission unit (U), and the server that performs clustering of the transmission unit (U) may perform clustering for each transmission unit (U). You can check the conditions of the pre-set delivery unit (U) as the basis for judging whether or not to group each product into the corresponding delivery unit (U).

1) Volume conditions (60L):

This is the limit volume (60L) of one transport unit (U) set by the user before clustering, and the corresponding products in the direction away from the transport center (C) to fill each transport unit (U) up to the limit volume (60L). can be grouped into the same transmission unit (U).

2) Distance condition (Lmax):

It is the limit distance (Lmax) from the delivery center (C) set by the user before clustering, and in particular, the distance between products classified as one delivery unit (U) in the lower priority delivery unit (U) set in the latter half of clustering, More specifically, if the distance from the convergence center (C) exceeds the preset limit distance (Lmax), 1) even if the volume condition (60L) is satisfied, the limit distance from the convergence center (C) is independent of the volume condition ( Products exceeding Lmax) may not be bundled into one delivery unit (U). Throughout this specification, the distance from the delivery center (C) to each product or the distance between neighboring products refers to the shortest distance and may correspond to the Euclidean distance.

For reference, with respect to 1) volume condition (60L) and 2) distance condition (Lmax), in clustering according to an embodiment of the present invention, both 1) volume condition (60L) and 2) distance condition (Lmax) are satisfied. Only products that are classified are not classified into the same delivery unit (U). For example, in various embodiments of the present invention, each delivery unit (U) in the finally classified delivery unit (U) is 1) volume condition Only (60L) may be satisfied and 2) the distance condition (Lmax) may not be satisfied, or alternatively, 2) only the distance condition (Lmax) may be satisfied and 1) the volume condition (60L) may not be satisfied, or otherwise. , both 1) volume condition (60L) and 2) distance condition (Lmax) may not be satisfied.

This is the case where products that 1) satisfy only the volume condition (60L) and 2) do not satisfy the distance condition (Lmax) are classified into the same forwarding unit (U), for example, the subordinate forwarding unit (U) set in the latter half of clustering. Products that exceed the limit distance (Lmax) from the delivery center (C) may also be classified into the same delivery unit (U). In this case, the server can request the server operator to allocate transportation resources suitable for long-distance delivery of relatively lightweight products.

2) This is a case where products that satisfy only the distance condition (Lmax) and 1) do not satisfy the volume condition (60L) are classified into the same delivery unit (U), and the difference between the volume of the subordinate delivery unit (U) and the limit volume (60L) Based on the comparison, for a collection unit (U) with a relatively small volume that is deemed inefficient to be classified as a single delivery unit (U), the delivery unit (U) is released, and the products of the released delivery unit (U) are released. By classifying them into other nearest neighboring conveying units (U), the volume of the corresponding conveying unit (U) may exceed the limit volume (60L) (e.g., conveying unit U1 in Figure 5A), in which case, 2) The distance condition (Lmax) may be satisfied, but 1) the volume condition (60L) may not be satisfied, and both 1) the volume condition (60L) and 2) the distance condition (Lmax) may not be satisfied.

As such, in clustering according to an embodiment of the present invention, corresponding products can be classified into the same delivery unit (U) depending on 1) whether the volume condition (60L) is satisfied, and 2) whether the distance condition (Lmax) is satisfied. For example, products that satisfy both 1) volume condition (60L) and 2) distance condition (Lmax) may be classified as the same delivery unit (U), but 1) volume condition ( Even in the case of products that do not satisfy 60L) or 2) products that do not satisfy the distance condition (Lmax), the distribution of transportation resources may be different or clustering in other neighboring delivery units (U) may result in the same delivery unit. It can be classified as (U).

In one embodiment of the present invention, 1) even if the volume condition (60L) is satisfied, 2) the distance condition (Lmax) is not satisfied, or 2) the distance condition (Lmax) is not satisfied, and 2) the distance condition (Lmax) is not satisfied. When the forwarding unit (U) is determined according to the above, especially for the lower priority forwarding unit (U) set in the latter part of clustering, each forwarding unit (U) is calculated through a separate calculation to be 1) equal to or less than the limit volume (60L) set in the volume condition. Depending on the difference, for example, if the volume of the junior forwarding unit (U) is relatively small enough to be considered inefficient to be grouped into one forwarding unit (U), and the difference from the limit volume (60L) is large, the subordinate forwarding unit (U) Rather than maintaining U), the setting of the subordinate forwarding unit (U) may be canceled and instead, each product may be distributed to the forwarding unit (U) to which the other nearest neighboring products belong (see arrow in Figure 5a). In this case, despite exceeding the limit volume (60L) set as the consignment unit (U), the excess volume exceeding the limit volume (60L) may be accommodated through distribution of transportation resources (see consignment unit U1 in Figure 5a) .

In various embodiments of the present invention, 1) the volume condition (60L) is satisfied, but 2) the distance condition (Lmax) is not satisfied, or 2) the distance condition (Lmax) is not satisfied, and 2) the distance condition (Lmax) is satisfied. If the forwarding unit (U) is determined according to this, especially for the subordinate forwarding unit (U) set in the latter half of clustering, the subordinate forwarding unit (U) itself is maintained, but forms the center of the corresponding subordinate forwarding unit (U). Only products subject to forwarding (specific products) that exceed the limit distance from the forwarding center (C) are excluded from the subordinate forwarding unit (U) and transferred to the forwarding unit (U) to which other neighboring products closest to the relevant specific product belong. can be distributed (see arrow in Figure 5a).

In various embodiments of the present invention, 1) the volume condition (60L) is satisfied, but 2) the distance condition (Lmax) is not satisfied, or 2) the distance condition (Lmax) is not satisfied, and 2) the distance condition (Lmax) is satisfied. If the forwarding unit (U) is determined according to , the forwarding unit (U) may be maintained (e.g., see forwarding unit U2 in Figure 5A, requesting distribution of transportation resources corresponding to short-distance forwarding of light volume), Alternatively, the forwarding unit (U) itself may be maintained, but only specific products exceeding the limit distance (Lmax) may be distributed to the forwarding unit (U) to which other nearest neighboring products belong (see arrow in Figure 5a), and the forwarding unit Since the volume of the collection unit (U) is relatively small to the extent that it is inefficient to bundle it into (U), the collection unit (U) with a large difference from the limit volume is released and each product belonging to the released collection unit (U) is selected. It can be distributed to the transport unit (U) to which other neighboring products belong (see arrow in Figure 5a, which requests distribution of transport resources considering the excess volume exceeding the limit volume of 60L of transport unit U), or Figure 5b Reclassification of the delivery unit (U) as shown in may be performed.

As shown in Figure 5b, 2) when the forwarding unit (U) is determined according to the distance condition (Lmax), the forwarding unit (U) set in the previous step, including the corresponding forwarding unit (U), is released, and forwarding is performed Reclassification of the collection unit (U) can be performed by performing the classification of the unit (U) again, and by proceeding with a different process from the initial classification of the collection unit (U), the subordinate collection unit (U) set in the latter half of the clustering ) can solve the problem of including lightweight products scattered over a wide area. For example, as shown in FIGS. 5A and 5B, in the initial classification, 2) products (specific products) that exceed the limit distance (Lmax) from the delivery center (C) according to the distance condition (Lmax) are subjected to new delivery. You can set it as the center (C1, specific delivery center C1), and set the delivery unit (U1, specific delivery unit U1) with the newly set delivery center (C1, specific delivery center C1) as the center, and set the delivery unit (U1, specific delivery unit U1) as the center. ), that is, products belonging to a specific delivery unit (U1) are excluded, and for other products, the product with the largest volume is set as the delivery center (C2), substantially the same as the initial classification, and the set delivery center You can set a delivery unit (U2) with (C2) as the center. At this time, in various embodiments of the present invention, in the initial classification as shown in FIG. 5A, 2) products exceeding the limit distance (Lmax) from the delivery center (C) according to the distance condition (Lmax) are recognized as specific products; , in the reclassification as shown in Figure 5b, when reclassification of the delivery unit (U) is performed by setting a specific recognized product as a specific delivery center (C1) and setting a specific delivery unit (U1), the first In classification, specific products may be recognized as multiple. For example, after terminating the classification of all products subject to mailing in the initial classification, by performing reclassification, the maximum number of specific products that can come out of the initial classification can be selected. It can be induced to calculate (pre-processing clustering to calculate specific products), and by applying sequential clustering based on the size relationship among the maximum number of specific products, the primary collection center (C1) is determined from among the maximum number of specific products. It is set as the product with the largest volume, and adjacent products can be set as the same primary delivery unit (U1) until the limit volume is filled in the direction away from the set primary delivery center (C1). Among the remaining specific products excluded, the specific product with the next highest volume is set as the secondary delivery center (C2), and adjacent products are placed in the same secondary delivery center until the limit volume is filled in the direction away from the set secondary delivery center (C2). Following the method of setting the delivery unit (U2), the delivery center (C) can be set according to the size relationship of the volume among the maximum number of specific products, and after all specific products are classified, specific products from the previous stage are placed as the delivery center. For other products excluding the delivery unit (U) with (C) as the center, the product with the largest volume is set as the 3rd delivery center (C) according to the size relationship of volume, substantially the same as in the first classification. You can then continue setting up the 3rd delivery unit (U) with the corresponding delivery center (C) as the center.

For reference, throughout this specification, setting a transmission unit (U) with the transmission center (C) as the center means that the transmission center (C) corresponds to the central position of the transmission unit (U) in terms of location. This does not mean that, with the delivery center (C) as the origin, each product is grouped into the same delivery unit (U) based on the short distance from the delivery center (C) (advancing from the short distance to the far distance). , In this sense, the collection center (C) can be said to correspond to the center of clustering.

<Selection of products based on location>

Hereinafter, with reference to FIGS. 6 to 8, selection of a product based on the location of the purchaser will be described.

FIG. 6 is a diagram illustrating a first selection path in which product selection is initiated from information on the product itself, regardless of the location of the purchaser.

FIG. 7 is a diagram illustrating a second selection path in which product selection is initiated from the purchaser's location information.

FIG. 8 is a diagram schematically showing a server in which the product list (link information) provided in the second selection path shown in FIG. 7 is stored in the form of a web document for each purchaser.

FIG. 9 shows a diagram illustrating another second selection path in which product selection begins from the purchaser location for each theme product.

Referring to FIG. 7, in one embodiment of the present invention, a server that performs a commerce relay service between a buyer and an acquirer may provide a relay service for one commerce zone, and in this case, the server can provide one commerce relay service. Based on the location information of multiple buyers located in the commercial area, product information of each buyer located in the commercial area can be provided along with a street view of the corresponding commercial area.

In one embodiment of the present invention, the server can transmit a file (web document) in the form of a document such as HTML (HyperText Markup Language) or XML (Extensible Markup Language) to the purchaser's terminal connected to the server, and the purchaser The location information of the purchaser and the product information handled by each purchaser can be matched and displayed on the street view of the commercial area relayed by the server through an application such as a web browser running on the terminal of . In one embodiment of the present invention, by matching and displaying product information of purchasers on the street view of the commerce zone relayed by the server, even if you do not visit the commerce zone relayed by the server directly, you can purchase a virtual view of the commerce zone relayed by the server. It is possible to provide a shopping experience, and by providing a virtual shopping experience to remote buyers far from the commercial area, for example, overseas buyers, the server relays a commercial area, such as a prominent market such as Namdaemun Market. Interest in the commercial area can be increased, and the virtual shopping experience provided online can lead to increased sales in the relevant commercial area. In addition to increasing online sales through the server's commerce relay, it is also possible to increase interest in the commercial area. By encouraging direct visits, it can lead to an increase in offline sales.

In one embodiment of the present invention, the server is capable of transmitting a web document such as HTML to the purchaser's terminal connected to the server, and can transmit a web document such as HTML to the purchaser's terminal through an application such as a web browser running on the purchaser's terminal. , can provide a street view of the commercial district. At this time, the server generates a web element (e.g., an image hyperlink such as a direction sign or a text hyperlink such as a string, hereinafter referred to as a web element) displayed on a specific area or street view set on the street view of the commercial district. Depending on the click event (onclick or mouseover), the web document connected to the hyperlink is referred to, the link information of the web document is referred to, and a new street view with the few points moved is provided or the product connected to the hyperlink is provided. Information can be displayed on the buyer's terminal.

For example, in one embodiment of the present invention, the server refers to web documents stored in the server and multimedia such as images or videos that provide street views as lower layers according to priorities such as z-index. The content can be executed, and it overlaps with the street view as a top layer and contains web elements (image hyperlinks such as direction signs or text hyperlinks such as strings) that are displayed in a specific area of the street view or overlaid on the street view. The layers are arranged to overlap on the street view of the lower layer, and the street view is updated according to a new view point according to a click event that occurs in a specific area or web element set in the upper layer, or the purchaser's products displayed in the street view. The street view can be displayed overlapping on the displayed tab or displayed on a new tab.

In one embodiment of the present invention, while providing a street view of a commercial area containing images of a plurality of purchasers, the information is displayed from a specific area set on the image of each purchaser or a web element displayed overlapping on the image of each purchaser. Depending on the click event (onclick or mouseover) that occurs, a list of products from the corresponding purchaser can be provided, and this product list includes multiple product images from the purchaser according to the click event of the web element displayed overlapping on the image of the purchaser. It may include, and may be displayed on the buyer's terminal by a server that references a web document linked to a web element corresponding to the purchaser.

In one embodiment of the present invention, the server may refer to a web document linked to the image of the purchaser in which a click event occurred and provide a list of products of the corresponding purchaser. Referring to FIG. 8, the server may store a web document in which a product list (link information) of each purchaser is inserted corresponding to each purchaser. The server may store a web document in which a product list (link information) is inserted for each purchaser, and each purchaser can create product images (images) through editing the web document in which its product list (link information) is inserted. Modifications such as the arrangement of hyperlinks, addition of new product images, or deletion of old or low-selling product images can be made more freely. In one embodiment of the present invention, a web document containing link information for brief information on a product provided in a second selection route initiated from location information of a purchaser, such as a street view, may be stored, and the web document may be stored. Modification authority may be granted to each purchaser, depending on the account type logged in to the server (see FIG. 8).

Referring to FIG. 7, in one embodiment of the present invention, the server refers to a web document in which link information of a product list is inserted, and an image hyperlink expressed as a product image as a product list is divided into a plurality of rows and columns. A web arranged in a table format for each cell can be displayed on the buyer's terminal. For example, in one embodiment of the present invention, the product list is a list of products of a purchaser for which a click event occurred on the street view, and two or more products may be displayed together on one window, and the product list of the present invention In various embodiments, each product image (image hyperlink expressed as a product image) may be displayed in each divided area of one window, such as in a table format, on the window.

In one embodiment of the present invention, the server may provide product information linked to each image hyperlink according to a click event for the image hyperlink represented by each product image, for example, The server may refer to a web document in which each product information (link information) linked to the hyperlink represented by each product image is inserted, and display the corresponding product information on the buyer's terminal. In other words, the server sends the corresponding product information linked to each image hyperlink to the buyer's terminal according to a click event (onclick or mouseover) generated from the image hyperlink represented by each product image on the product list of the corresponding purchaser. It can be displayed on the screen.

As will be described later, in one embodiment of the present invention, the products included in the product list of each buyer include a first group of products for which detailed information is registered according to the type of product information registered in the server; It can be classified as a second group of products in which only brief information, not detailed information, is registered. Below, the products of the first and second groups will be described in more detail.

Referring to FIG. 7, in one embodiment of the present invention, the products for which information is provided by matching the location information of the purchaser on the street view of the commercial transaction area are the first and second groups with different types of information registered on the server. may include products. As the first group product information, the server may store detailed information about the first group products, and as the second group product information, the server may store detailed information about the second group products. Instead, only brief information about the second group of products may be stored.

In one embodiment of the present invention, the server (server database) may store a plurality of product information registered from a plurality of different purchasers, and the product information may include a first group including detailed information. It may include product information and a second group of product information that includes only brief information rather than detailed product information.

Throughout this specification, the first group of products refers to commercial products relayed by the server, detailed information of which is registered on the server, and detailed information is displayed or at least a link to detailed product information such as button type or bookmark type is provided. It may refer to a product for which a purchase request can be made through the displayed web, and as described above, a purchase button connected to the purchase procedure may be provided on the web related to detailed information according to a click event as a purchase request.

In one embodiment of the present invention, the first group of products is selected from a system in which product selection is initiated from information on the product itself, regardless of the location information of the purchaser, for example, without going through a street view in which the location of the purchaser is displayed. Product selection may be made by a first selection path (see FIG. 6) or, alternatively, by a second selection path in which product selection is initiated from the location information of the purchaser (e.g., a street view displaying the location of the purchaser). There is (see Figure 7). Here, the first and second selections do not mean priority of selection, and product selection is initiated from information on the product itself regardless of the location of the purchaser (first selection, see FIG. 6) or location information of the purchaser (e.g. For example, there may be a difference in the selection path starting from the street view showing the location of the purchaser (second selection, see Figure 7), and the first group of products has two selection paths, the first and second selection. A purchase can be made after being selected according to .

In one embodiment of the present invention, the selection of a product involves two different selection paths: a first selection initiated from information on the product itself that is not based on the location information of the purchaser, and a second selection initiated from the location information of the purchaser. It can be done along, and the first selection path and the second selection path will be described in more detail later.

Throughout this specification, the second product refers to a commercial product relayed by a server in which only brief information, not product detailed information, is registered on the server, and brief information about the product is displayed or product information such as button type or bookmark type is registered on the server. It may refer to a product for which a purchase request can be made through the web where a link to brief information is displayed. As described above, on the web related to brief information of the product, a purchase procedure may be initiated according to a click event as a purchase request by the user. A purchase button may be provided. In one embodiment of the present invention, the second product is a first selection path ( The purchase may be made after being selected through the second selection path (see FIG. 7), in which product selection is initiated from the purchaser's location information, rather than by (see FIG. 6).

Referring to Figure 6, in one embodiment of the present invention, product information may be arranged on the web displayed on the buyer's terminal connected to the server, regardless of the location information of the purchaser, for example, the buyer's inquiry text Multiple product information retrieved from the server (server database) according to the (search term) is in a table format (or list format) with cells divided along rows and/or columns, or with multiple frames divided into one window. It can be arranged in the form of a frame set, and each divided cell or frame can be linked to a new web containing detailed information of the corresponding product according to a click event. For example, in one embodiment of the present invention, a web provided by a web document received from the server may be implemented on the purchaser's terminal connected to the server, and a form implemented on the web and the server side may be implemented. In cases where a purchase request is made for a product selected by the buyer from among products provided in response to the buyer's search term obtained through CGI (common gateway interface), a request to purchase a product may be made not based on the location information of the purchaser. There may be a case, and this case can be referred to as the first selection path in which product selection begins from information on the product itself, regardless of the location of the purchaser. As such, in one embodiment of the present invention, regardless of the location information of the purchaser, the location information of the purchaser, that is, the street view displaying the location of the purchaser, may not be provided on the first selection route initiated from the information of the product itself. And the selection of a product can be initiated from information on the product itself, regardless of the location information of the purchaser.

The first group of products may be selected along a first selection path (see FIG. 6) initiated from information on the product itself, and alternatively, a second selection path initiated from location information of the purchaser (see FIG. 7). It may be selected according to . For example, the first product may be selected along a second selection path where detailed information about the product is confirmed through a hyperlink connected to the product list of the purchaser on a street view displaying the location of the purchaser. At this time, detailed information on the first group of products may be stored in the server, and the detailed information on the first group of products is provided as the same content in the first and second selection paths (see FIGS. 6 and 7). It can be. For example, a list of products searched according to the buyer's query text (search term) along the first selection path may be linked with link information leading to detailed information about each product (see FIG. 6), 2 Link information leading to detailed information about each product may be connected to the list of products linked to each buyer location along the selection path (see Figure 7), and the products obtained through the first and second selection paths may be connected. Detailed information may be linked to the same content (same content - or a web document into which link information thereof is inserted), and detailed information of the same product is provided regardless of the first and second selection paths (see FIGS. 6 and 7). can be provided.

The second group of products cannot be selected along the first selection path (see FIG. 6) starting from information on the product itself, but from the location information of the purchaser (for example, a street view displaying the location of the purchaser). The selection may be made along the disclosed second selection path (see FIG. 7). For example, the second product may be selected along a second selection path (see FIG. 7) that confirms brief information about the product through a hyperlink connected to the product list of the purchaser on a street view displaying the location of the purchaser. . At this time, the server may store brief information on the products of the second group, rather than detailed information on the products of the second group, and the brief information on the products of the second group may be stored in the image hyperlink of the purchaser displayed on the street view. After the product list of the purchaser is output through the link, the image of the individual product displayed in the product list can be displayed on the buyer's terminal through a hyperlink. For example, the brief information includes the image of the product and Together, essential information for purchasing, such as price information or the phone number of the purchaser, can be provided (see Figure 7).

In one embodiment of the present invention, the server can display detailed information on products registered from purchasers on the web and relay commercial transactions for the products. At this time, there may be inconvenience for the purchaser to register detailed information such as the manufacturer of the product and the material of the product, including the image of the product, on the server. Additionally, when registering a new product, the purchaser's information on the product may be inconvenient. Because of the difficulty in determining preferences and the inconvenience of registering a new product as described above, people may hesitate to register a new product. In one embodiment of the present invention, unlike the relay for the first group of products, the second group of products only needs to register brief information, for example, essential information for purchase, along with the product image, and also New products can be easily registered by modifying the dedicated web or web document assigned to the purchaser (editing rights are granted depending on the type of account logged in to the server, see Figure 8), and the first group of products is registered on the server. In that it is possible to gauge the buyer's preference before registering the product, for example, if sales of the product increase through the second selection route (see Figure 7) initiated from the street view as a purchaser, the first group Before registering with the detailed information of the product as a product of the first group, such as registering on the server with the detailed information of the product as a product of the first group (see FIG. 6, first selection path), it is possible to facilitate transaction of the product. (First selection path, see FIG. 6), sale of the product can be started through a second selection route such as street view (second selection route, see FIG. 7) with only brief information about the product. In addition, in the case of Group 1 products that are registered with detailed product information on the server's web, a more conservative pricing policy can be maintained rather than a flexible pricing policy, and price changes may affect the reliability of the server operator. However, in the case of the second group of products provided through a hyperlink to the buyer's image on the street view, only brief information essential for purchase is provided, and more flexible information that can be changed at any time through the web or web documents managed by the buyer A pricing policy is possible, and before registering a product in the first group on the server, it is possible to estimate the appropriate price for the product in the second group by considering sales volume.

In one embodiment of the present invention, the classification of the first and second group products is done by inducing registration as the second group product before registration as the first group product, and as a result, the first group product Since registration can be encouraged, the server operator can also promote product registration despite the inconvenience of registration.

Referring to FIG. 8, a web document containing a product list (link information) linked through a hyperlink of the purchaser image displayed on the street view can be individually assigned to each purchaser, and the web document assigned to the purchaser is Through editing, modifications such as the arrangement of product images (image hyperlinks), addition of new product images, or deletion of old or low-selling product images can be made more freely.

In one embodiment of the present invention, authentication such as member registration, member login, etc. can be performed through the form of a web document received at the purchaser's terminal connected to the server and the CGI (common gateway interface) on the server side, where , members can include both purchaser members and buyer members on both sides of the server relaying commercial transactions. For example, members certified as purchaser members can more flexibly display their products on the web through a dedicated web or web document assigned to each purchaser member, and can display products in groups 1 and 2 as products. You can register and update on the server with your information.

Referring to FIG. 9, as another embodiment of the second selection path in which product selection is initiated from the location information of the purchaser, the purchaser for each theme product is displayed on the buyer's terminal along with an overall view of the commerce area where the server relays commerce. For example, a group of purchasers handling first theme products selectively selected according to the buyer's request can be collectively displayed. For example, the first theme can be displayed on the entire view of the commercial area. A group of purchasers that handle products can be displayed in one place. Additionally, the server may collectively display another group of purchasers handling second theme products selectively selected according to the buyer's request on the entire view of the commercial area.

For example, the server can divide the window and display a web on the buyer's terminal that provides an overall view of the commercial area in the upper frame and displays a web with radio buttons from which various themes can be selected in the lower frame. At this time, according to the alternative selection of different theme products, for example, first theme products and second theme products, which are alternatively displayed with a radio button, the server is a group that handles the selected first theme product. The purchaser can be displayed overlapping with the overall view on the upper frame.

For example, in one embodiment of the present invention, the server refers to a web document stored on the server and displays multi-media content, such as an image providing an overall view of the commercial area, in the upper frame dividing the circular diagram. It can be executed, and the lower frame can display selectable radio buttons for different theme products, and link information connected to each radio button for different theme products can be inserted in the web document.

Depending on the selection input of the radio button, the server can execute multimedia content such as images that provide an overall view as a lower layer according to priority such as z-index, and web documents linked to the selected radio button as the upper layer. The upper layer, which includes web elements (image hyperlinks such as dot marks) that display the location of the purchaser handling the corresponding theme product, is arranged to overlap the entire view of the lower layer, and in the image hyperlink of the upper layer Depending on the click event that occurs, the buyer's product (or theme product) corresponding to the buyer's location displayed in the overall view is displayed overlapping on the tab (window) where the overall view is displayed, or displayed on a new tab (window). You can. For example, in one embodiment of the present invention, the server contains a web document for implementing a web in which the location of a group of purchasers handling each theme product is indicated by an image hyperlink such as a dot. It may be stored. In addition, depending on the click event of the image hyperlink indicating the location of each purchaser, the product list of each purchaser, that is, the list of products provided by clicking the image hyperlink of the purchaser displayed in the street view, is output in the same way or is displayed in the same way as each purchaser. A theme product list that arranges only the theme products selected among the products in the form of image hyperlinks may be output.

<Private settings to prevent imitation of small fashion items>

Hereinafter, with reference to FIGS. 10 to 15C, private settings for preventing imitation of small fashion items will be described.

FIG. 10 shows a diagram showing an embodiment of a web implemented for relaying commercial transactions for products set to be private.

FIG. 11 shows a diagram illustrating the one-to-one relationship between a purchaser and a purchaser for disclosing a product set as private.

12A to 12C are diagrams illustrating different classes that can be classified according to external characteristics in accessories such as earrings.

Figure 13 shows a diagram to explain an artificial intelligence neural network model that implements a regression algorithm that analyzes the correlation between multiple attributes and class labels as characteristics of the appearance of a private product.

Figure 14 shows a diagram to explain a convolutional neural network model that predicts the probability that a non-public product belongs to a non-public product using the product image itself as input.

15A to 15C are examples of filters included in the convolutional neural network model, and a high value is applied along the appearance features of each class to generate a feature map that can extract the appearance features of each class belonging to the same product category. There are diagrams showing exemplary filters with weights of .

Referring to FIG. 10, according to one embodiment of the present invention, the server sets the image of the product itself as private so as not to disclose it, and instead of the image of the product, discloses a similar image to provide a rough image of the product. It is possible to relay commerce for products that are privately set so that they can be identified. For example, for private products, the server does not disclose the product image itself to general members based on the one-to-one relationship between the purchaser (purchase member) of the product and the purchaser (see Figure 11), but provides similar images. is disclosed, targeting buyers who have a relatively high rating in the one-to-one relationship between the purchaser (purchase member) and the buyer (see Figure 11), for example, buyers who have built up trust with the purchaser and have relatively little concern about product duplication. You can disclose the image of the product itself.

Referring to FIG. 11, in one embodiment of the present invention, the image of the product is disclosed according to a one-to-one relationship between the purchaser (purchase member) of the product and the purchaser, thereby replicating the product, especially the purchase value of the product as the appearance itself. In the case of small fashion items such as fashion items where price can be determined or accessories that can be copied quickly, product duplication can be blocked, and buyers can prevent the emergence of competition due to duplicate products and the resulting decrease in sales. there is.

In one embodiment of the present invention, as a prerequisite for disclosure of private products, the one-to-one relationship between the buyer and the purchaser is not based on objective data such as membership registration or member authentication, but on the consent of both parties between the purchaser and the buyer. It means a relative relationship established by, for example, a one-to-one relationship between a purchaser and a buyer stored on the server based on agreement between the purchaser and the buyer, rather than input of objective data and authentication from the server for the entered data. It can mean. For example, the one-to-one relationship may be displayed on the web in different expressions such as first-degree relative, neighbor, or friend, depending on the relative relationship between the purchaser and the purchaser, but their one-to-one relationship may be graded according to the relative relationship. For example, the level of this one-to-one relationship can be registered on the server by mutual consent between the specified purchaser and the specified buyer, and the level of this one-to-one relationship is between the specified purchaser (purchase member) and the specified buyer. It is set in a one-to-one relationship between (purchasing members), and is set objectively for a specific member itself, such as an excellent member registered on the server, that is, based on objective data such as the member's purchase data. It is different from the grade given. In one embodiment of the present invention, whether or not a private product is disclosed is different from whether the member is registered on the server side or whether the member is authenticated, and in the present invention, general member buyers are classified according to a one-to-one relationship with the purchaser of the product. Images of private products are not disclosed to other parties, and images of private products can only be revealed to members who have a high level of one-to-one relationship with the purchaser of the product.

In one embodiment of the present invention, a purchaser who has registered a private product can prevent the product from being quickly copied and forming a competitive relationship with the product by disclosing its appearance, but the disclosure of the product image is within a limited personal range. By doing so, sales of the product may be limited. Accordingly, in one embodiment of the present invention, even for products set as private, a similar image is provided to general members along with a notice to the effect that disclosure is determined based on the one-to-one relationship with the purchaser of the product. Even if you are a general member who cannot check the image of the undisclosed product, you can purchase the undisclosed product through one-on-one contact with the purchaser of the product in the future and avoid the competitive relationship caused by product duplication in the retail market and create an advantageous market such as a monopoly. can be formed. For example, in one embodiment of the present invention, the server may relay product transactions between a buyer classified as a wholesaler and a buyer classified as a retailer, and throughout this specification, the buyer refers to a business operator corresponding to a retailer rather than an individual member. Since you can be a member, you can promote the purchase of undisclosed products through one-on-one contact with the purchaser of the undisclosed product for the purpose of handling products with excellent marketability through undisclosed products. Therefore, in the present invention, even for products set as undisclosed, Through the disclosure of similar images, it is possible to understand the general outline of the product and induce one-on-one contact with the purchaser of the product.

In one embodiment of the present invention, for a private product, the probability of belonging to different classes classified within the same product category is calculated as a similar image that can be made public to general members, and the probability of belonging to the same class as the corresponding product is calculated. Together with this relatively high typical image of the same class, the calculated probability can be revealed together. In one embodiment of the present invention, classes are classified according to the appearance of the product within the product category to which the product belongs, the probability of belonging to the classified class is calculated, and a class with a relatively high probability or a typical image of the corresponding class is disclosed. can do.

In one embodiment of the present invention, the product image itself is set to be private, and the private product in which similar images are made public is a fashion item whose purchase value can be determined by the appearance of the product, such as accessories that can be copied in a relatively short period of time. It may correspond to a small fashion item, and may also correspond to a fashion item that can be classified into a class in which multiple products belonging to the same product category share some external characteristics (attributes). For example, in accessories such as earrings, stud or button type earrings (FIG. 12a) that stick to the attachment location (ear) and hang downward from the attachment location (ear) to swing. Drop or dangle type earrings (Figure 12b) that hang long in the form of a line, and rings that hang long in the form of a closed loop such as a circle, triangle, etc. so that they hang down from the attachment point (ear) and swing. They can be classified as ring or hoop type earrings (FIG. 12c).

In one embodiment of the present invention, a mathematical model or a neural network model of artificial intelligence is used to calculate the probability that a product set as private belongs to multiple classes classified within the same category, for example, classified with a relatively high probability. A typical image of a given class can be displayed on the web as an image similar to the corresponding private product. At this time, multiple classes belonging to the same product category include stud or button type earrings (Figure 12a), drop or dangle type earrings (Figure 12b), and rings. Alternatively, hoop-type earrings (FIG. 12c) are encoded as integers such as class 0, class 1, or class 2, or are encoded into a mathematical model or neural network model of artificial intelligence through one-hot encoding. Can be encoded.

In one embodiment of the present invention, the server inputs (input vector) a number of attributes (numerical shape factors such as length and width) as characteristics of the appearance of products belonging to the same product category (regression algorithm, Fig. 13) or a mathematical model or artificial intelligence neural network model that takes the product image itself as input and can predict class labels (convolutional neural network, see Figure 14), so that private products are classified within the same product category. The probability of belonging to each class can be calculated.

Referring to FIG. 13, in one embodiment of the present invention, the appearance of a private product includes a number of attributes (numerical shape factors such as length and width) and class labels (e.g., different types of earrings). ) can be used to analyze the correlation between the two, and this regression algorithm can be implemented as a neural network model of artificial intelligence. For example, in the regression algorithm, the probability that the product belongs to each class can be predicted from a linear combination of the multiple attributes (x) of the product's appearance. For example, the probability that the product belongs to each class can be predicted from the linear combination of the multiple attributes (x). The value of the weight (w) added to each attribute in the combination can be calculated through learning a neural network model. In one embodiment of the present invention, the linear combination of the multiple attributes (x) may include each attribute and a combination of each attribute, for example, the product of each attribute itself and arbitrarily It may include the product of two or more selected attributes.

More specifically, the artificial intelligence neural network model uses an input vector (x) in which a number of attributes (numerical shape factors such as length and width) of the appearance of products belonging to the same product category are arranged in an array, and learning Linear combination of the learned weight vectors (w, or learned weight matrices) with an activation function, for example, the sigmoid function in binary classification or the softmax function in multiclassification. The applied output can be output as a one-dimensional array in the form of a one-hot vector that displays the probability of belonging to each class at each column position.

Referring to Figure 13, in one embodiment of the present invention, the artificial intelligence neural network model for calculating the class to which a private product is likely to belong includes a number of attributes (numerical properties such as length and width) of the product's appearance. Shape factor) is a linear combination of an input vector (x) and a learned weight vector (w, or learned weight matrix) arranged in an array, or a single layer of neurons containing an activation function applied to their linear combination. A single-layer neural network model including an input layer (x) to which the input vector is applied, a hidden layer (z) fully connected to the input layer (x), and an output layer (z) fully connected to the hidden layer (z) A deep neural network model including multiple layers of neurons, such as y), can be applied. In one embodiment of the present invention, the learned weight vector (or learned weight matrix) is a pre-prepared learning data set, and the class label among the classes classified within the same product category is the attribute (length) of the product given as the target. By applying a backpropagation algorithm that calculates the partial derivative function to minimize the error or cost function between the predicted class label as the output of a neural network model with numerical shape factors such as width) as the input vector (x) and the given target label, The weight vector (w, or weight matrix) can be updated every epoch. For example, in one embodiment of the present invention, it is updated for each training data (stochastic gradient descent) or Alternatively, the learned weight vector (w, or weight matrix) can be obtained by updating each piece of training data (minibatch gradient descent).

In one embodiment of the present invention, the artificial intelligence neural network model for calculating the class to which a private product is likely to belong includes a number of attributes of the product's appearance (numerical shape factors such as length and width) and each Instead of a regression algorithm that analyzes the correlation between the probabilities of belonging to a class, a convolutional neural network model input as the product image itself can be applied.

Referring to FIG. 14, the convolutional neural network model alternates between a convolutional layer including a plurality of filters and a pooling layer (max pooling, average pooling) to reduce the size of the feature map calculated from the convolutional layer. It can include flattening, which unfolds the 2-dimensional matrix-type 3-dimensional feature map with the depth dimension into 1-dimensional form, and an output layer that is fully connected to the flattened 1-dimensional feature vector. Through the encompassing dense layer, the probability of belonging to each class can be output in the form of a one-hot vector.

For example, the convolution layer may include a plurality of filters, and a plurality of filters (each element of the filter) on R, G, B image data (pixel values as each element of the image data) as image data of a private product. As the weight w) is convoluted, a feature map with an increased depth dimension can be created while the two-dimensional size of the row decreases, and same padding (same padding) that pads with 0 along the periphery of the image data of a private product is used. By applying the same padding, the pixel value information according to each row and column position can be prevented from being biased depending on the position. In this case, the row and column size of the feature map can be maintained. In contrast, image data of private products can be maintained. Valid padding can also be applied to generate the feature map itself, in which case the row size of the feature map may be reduced.

For example, in the convolution layer, the filter is moved by the number of pixels corresponding to the preset stride on the two-dimensional image data of the private product, and the pixel values of the corresponding row and column positions and the weight of the filter are linearly adjusted. A feature map can be created by calculating a combination, for example, by applying an activation function (ReLu) to a linear combination of the pixel value of the image data of a private product and the weight (w) of the filter. can be created. In one embodiment of the present invention, a filter having a size of 3x3 or 5x5 may be applied in the convolution layer.

The pooling layer may use max pooling or average pooling to reduce the size of the feature map generated from the convolution layer. By applying the pooling layer rather than increasing the stride of the convolution layer, , better performance can be achieved while reducing the size of the feature map. The dense layer flattens the three-dimensional feature map with the two-dimensional and depth dimensions of the matrix into a one-dimensional feature vector, and performs a linear combination or linear combination of a plurality of elements forming the one-dimensional feature vector. By applying an activation function, the probability of belonging to each class classified within a product category, such as a private product, can be calculated. For example, the probability of belonging to each class is indicated by a different column position (index position). You can obtain a one-dimensional array in the form of a one-hot vector.

In one embodiment of the present invention, the weight (w) of the filter applied in the convolutional layer can be learned using a training data set containing a plurality of training data and a target class label given for each training data, e.g. For example, the weight of the filter can be updated every epoch by applying a backpropagation algorithm that calculates the partial derivative function to minimize the error or cost function between the predicted class label predicted from the convolutional neural network model and the given target label. You can.

Referring to FIGS. 15A to 15C, multiple filters can be applied in the convolution layer, and each filter is applied to each class to generate a feature map that can extract the appearance features of each class belonging to the same product category. Multiple filters with high weights can be applied according to the appearance characteristics of . For example, in order to classify button-type earrings (see Figure 12a) as multiple classes belonging to the same product category, a filter with a high weight (see Figure 15a) according to the button shape is applied to determine the button type. A feature map with a high output value in the corresponding area can be created through convolution with image data (input values) (Figure 15a). In addition, in order to classify drop-type earrings (see FIG. 12b), a filter with a high weight (see FIG. 15b) is applied along the vertical direction, and through convolution with drop-type image data (input value), the corresponding A feature map with high output values in the region can be generated (Figure 15b). Similarly, to classify ring-type earrings (FIG. 12c), a filter with a high weight (FIG. 15c) is applied along a closed loop surrounding the central hollow to obtain ring-type image data (input value). Through convolution with , a feature map with a high output value in the corresponding area can be generated (Figure 15c). However, according to one embodiment of the present invention, the convolutional neural network for image classification may include multiple convolutional layers, and the filter applied in the multiple convolutional layers may activate part of the image from the overall image of the private product. , The convolutional layer close to the input layer can recognize visual information or patterns, and can recognize higher-dimensional concepts as it goes from the input layer to the output layer. As explained previously, each filter corresponds to one class. Recognizing external appearance characteristics may correspond to one embodiment among various embodiments of the present invention.

As such, in one embodiment of the present invention, the probability that a private product belongs to each class can be predicted using a neural network model in which a regression algorithm is implemented or a convolutional neural network model, and the normalized probability value is 0.5 or more. The class predicted by the probability can be determined as the class to which the private product belongs, and the predicted probability can be made public along with a typical image of the corresponding class. For example, when the private product corresponds to a drop-type earring (see FIG. 12b), a neural network model or convolutional neural network model implementing a regression algorithm according to an embodiment of the present invention determines that the corresponding private product is a drop. The probability of belonging to the type class (see Figure 12b) can be predicted with a probability close to 1, while the probability of belonging to the button type class (see Figure 12a) or the ring type class (see Figure 12c) can be predicted as It can be predicted with a probability close to 0, and accordingly, the server does not disclose the image of the private product itself, but can provide a similar image as minimal information about the appearance of the private product. In this case, the private product As a similar image, a typical image belonging to a class predicted with a relatively high probability (normalized probability of 0.5 or more) can be disclosed as a similar image. In another embodiment of the present invention, a typical image belonging to a class predicted with a probability of 0.3 or more, rather than a normalized probability of 0.5 or more, may be disclosed as a similar image for the undisclosed product. If, as a similar image for a private product, only the class predicted with a normalized probability of 0.5 or more is disclosed as a similar image, only one similar image can be disclosed, and also the class predicted with a normalized probability of 0.5 or more If there is no, for example, in the case of products sharing the attributes of two or more classes, the probability of belonging to each class is unlikely to be calculated as a substantially normalized 0.5 or more, so in one embodiment of the present invention, the For products set to , typical images belonging to the class predicted with a normalized probability of 0.3 or higher can be released as similar images.

For example, the server may store typical images belonging to each class, and since these typical images clearly have appearance characteristics that classify each class, the neural network model clearly identifies the appearance characteristics of each product. It can be used as learning data for a neural network model so that it can be recognized clearly.

<Detection of trend changes due to sudden changes in buyer preferences>

Hereinafter, with reference to FIGS. 16 and 17, a series of processes of the server that detects trend changes due to sudden changes in the buyer's preferences will be described.

Figure 16 shows a diagram showing the extraction of index candidate products of the first and second groups according to rapid changes in buyer preferences.

Figure 17 is a diagram showing a series of processes of a server that outputs a trend indicator product image by inputting a group of indicator candidate product images.

Referring to Figures 16 and 17 together, in one embodiment of the present invention, the server periodically calculates the buyer's preference, and selects a group of products for which the buyer's preference changes rapidly from the calculated buyer's preference as an indicator candidate. It is possible to recognize products as products, select products with common characteristics among these index candidate products as trend index products, and disclose the trend index products to purchasers who register products on the server. In one embodiment of the present invention, it is a tendency that influences the buyer's preference, such as environmental factors such as seasonal changes or weather changes, changes in life trends such as disease epidemics, and fashion trends due to the popularity of movies and dramas. By identifying trends that influence preferences, trend indicator products can be released to respond sensitively to these changes. In one embodiment of the present invention, by disclosing a group of trend indicator products rather than disclosing the tendency itself that influences the buyer's preference, damage to the purchaser due to the server's erroneous judgment can be prevented, and also excessive computation by the server can be prevented. It can reduce the burden.

In one embodiment of the present invention, the server may calculate the buyer's preferences at regular intervals, such as daily or weekly. In one embodiment of the present invention, the buyer's preference may mean the purchase ranking in which the buyer's purchase was made, or may refer to the purchaser's purchase ranking for a comprehensive individual product considering the search ranking based on the buyer's inquiry along with the purchase ranking. It can be understood as a quantitative understanding of the interests of

The server may calculate a change in ranking between the current buyer's preference and the buyer's preference calculated immediately before based on the periodically calculated buyer's preference, and may cause a sudden change in the calculated ranking change, for example, a sudden change in ranking. A group of products that show a rise or a sharp decline in ranking can be recognized as indicator candidate products. In one embodiment of the present invention, the server can extract products that share common characteristics from a group of indicator candidate products as trend indicator products.

In one embodiment of the present invention, the server can extract the trend indicator product itself rather than extracting the common features of the trend indicator product itself, and accordingly, the server can extract common features among a group of indicator candidate products. You can extract products that share .

In one embodiment of the present invention, the purchase value of a product belonging to a fashion item can be determined by the appearance of the product itself, so by analyzing the external characteristics of a group of index candidate products, the external characteristics of these index candidate products are determined. Shared products can be extracted and selected as trend indicator products. For example, in one embodiment of the present invention, the trend indicator product may be selected on a product category basis, and like earrings, it is a fashion item whose purchase value is determined by its appearance itself, and the classification of the class is clear from its appearance characteristics. As a product category belonging to small fashion items (i.e., accessories), a trend indicator product can be selected from one product category called earrings.

For example, in the product category of earrings, based on external characteristics, there are stud or button type earrings that stick to the attachment location (ear) (see Figure 12a) and attachment location (ear). ), drop or dangle type earrings (see FIG. 12b) that hang down and swing in a line, and earrings such as circles, triangles, etc. that hang down from the attachment point (ear) and swing. They can be classified into different classes, such as ring-shaped or hoop-type earrings (see Figure 12c), which are long in the form of a loop, and a convolutional neural network trained to recognize each class classified in this way is used. Through this, products with shared appearance characteristics can be extracted from a group of indicator candidate products and selected as trend indicator products.

Referring to FIG. 14, the server takes image data of a group of index candidate products as input and generates a feature map or a feature vector in the form of an array in which the feature maps are one-dimensionally arranged. The server may include an artificial intelligence neural network model to generate a feature map or feature vector including the external features of the product, and in one embodiment of the present invention, it may include a convolutional neural network model. You can.

For example, the convolutional neural network model may alternately include a convolutional layer including a plurality of filters and a pooling layer (max pooling, average pooling) to reduce the size of the feature map calculated from the convolutional layer. Finally, flattening is performed to unfold the 2-dimensional and depth-dimensional 3-dimensional feature maps in the form of rows and columns into 1-dimensional form, and an output layer is fully connected to the flattened 1-dimensional feature vector. Through the encompassing dense layer, the probability of belonging to each class can be output in the form of a one-hot vector.

For example, the convolution layer may include a plurality of filters, and a plurality of filters (each element of the filter) may be added to the R, G, B image data (pixel value as each element of the image data) as the image data of the indicator candidate product. As the weight, w) is convolved, a feature map with an increased depth dimension can be created while reducing the two-dimensional size of the matrix, and same padding is padded with 0 around the image data of the indicator candidate product. By applying (same padding), it is possible to ensure that the pixel value information according to each row and column position is not biased depending on the location. In this case, the row and column size of the feature map can be maintained. In contrast, the image of a private product can be maintained. Valid padding, which creates a feature map from the data itself, can also be applied, in which case the row size of the feature map may be reduced.

For example, in the convolution layer, the filter is moved by the number of pixels corresponding to the preset stride on the two-dimensional image data of the private product, and the pixel values of the corresponding row and column positions and the weight of the filter (w ) can be calculated to create a feature map. For example, an activation function (ReLu function) is applied to the linear combination of the pixel value of the image data of the indicator candidate product and the weight (w) of the filter. You can create a feature map using this method. In one embodiment of the present invention, a filter having a size of 3x3 or 5x5 may be applied in the convolution layer.

The pooling layer may use max pooling or average pooling to reduce the size of the feature map generated from the convolution layer. By applying the pooling layer rather than increasing the stride of the convolution layer, , better performance can be achieved while reducing the size of the feature map. The dense layer flattens the three-dimensional feature map with the two-dimensional and depth dimensions of the matrix into a one-dimensional feature vector, and performs a linear combination or linear combination of a plurality of elements forming the one-dimensional feature vector. By applying an activation function, the probability of belonging to each class classified within the same product category as the indicator candidate product can be calculated, for example, one-hot where the probability of belonging to each class is displayed in different column positions. You can obtain a one-dimensional array in the form of a vector.

In one embodiment of the present invention, the weight (w) of the filter included in the convolution layer can be learned using a training data set containing a plurality of training data and a target class label given for each training data, For example, the weight of the filter is adjusted every epoch by applying a backpropagation algorithm that calculates the partial derivative function to minimize the error or cost function between the predicted class label predicted from the convolutional neural network model and the given target class label. It can be renewed.

Referring to FIGS. 15A to 15C, the convolution layer may include a plurality of filters, and each filter is configured to generate a feature map capable of recognizing the appearance characteristics of each class belonging to the same product category. Depending on the external characteristics of the class, it may have a high weight (w). For example, in order to classify button-type earrings (Figure 12a) as multiple classes belonging to the same product category, by applying a filter (Figure 15a) with a high weight (w) according to the shape of the button, the button Through convolution with image data (input values) of this type, a feature map with a high output value can be created in the corresponding area. In addition, in order to classify drop-type earrings (FIG. 12b), a filter with a high weight (w) (FIG. 15b) is applied along the vertical direction, through convolution with drop-type image data (input value). It is possible to create a feature map with high output values in that area. Similarly, in order to classify ring-type earrings (FIG. 12c), a filter with a high weight (w) (FIG. 15c) is applied along a closed loop surrounding the central hollow to obtain ring-type image data (FIG. Through convolution with the input value), a feature map with a high output value can be created in the corresponding area. However, according to one embodiment of the present invention, the convolutional neural network for image classification may include multiple convolutional layers, and the filter applied in the multiple convolutional layers may activate part of the image from the overall image of the indicator candidate product. Since the convolutional layer close to the input layer can recognize visual information or patterns, and can recognize higher-dimensional concepts as it goes from the input layer to the output layer, each filter has one class as explained previously. Recognizing the corresponding external feature may correspond to one embodiment among various embodiments of the present invention. However, in one embodiment of the present invention, rather than recognizing complex patterns, the server recognizes simplified patterns with relatively clear distinctions in external features, such as earrings, and can classify different classes within the product category of earrings, Considering the computational burden of the server, the number of convolution layers can be minimized, and accordingly, each filter included in the convolution layer can recognize the external characteristics of each class.

Referring to FIG. 16, in one embodiment of the present invention, the server periodically calculates the buyer's preference, and selects a product with a relatively large change in ranking corresponding to the difference between the current preference and the previous preference as an indicator candidate product. It can be recognized, and as the output of a convolution model that takes as input the image data of a group of collected indicator candidate products, the probability that the corresponding indicator candidate product belongs to each class can be calculated. At this time, among a group of indicator candidate products, different indicator candidate products with similar probabilities of belonging to each class calculated as the output of the convolution model can be selected as trend indicator products. For example, in one embodiment of the present invention, the output of the convolution model is in the form of a one-dimensional vector (feature vector) in which the probability of belonging to each class is indicated at the index (column position) corresponding to each class. (see Figure 17, in feature vector Y1: (y11, y12, y13) and feature vector Y2: (y21, y22, y23), the probability of belonging to class 0, y11, y21, belonging to class 1, respectively. Probability y12, y22, probability of belonging to class 2 y13, y23), in one embodiment of the present invention, among a group of indicator candidate products, the feature vectors (Y1, Y2, Figure 17) corresponding to the output of the convolution model are similar to each other. Other indicator candidate products can be selected as trend indicator products. For example, in one embodiment of the present invention, the probability of belonging to the class corresponding to each index may be displayed at each column position in the one-dimensional feature vector (Y1, Y2, Figure 17), and each class Different indicator candidate products with similar probabilities of belonging can be selected as trend indicator products. In one embodiment of the present invention, the server calculates the Euclidean distance (S) between one-dimensional feature vectors (Y1, Y2, Figure 17) calculated as the output of a convolution model that inputs image data of a group of indicator candidate products. By calculating, it is possible to determine whether appearance features are shared between different indicator candidate products, for example, the Euclidean distance (S) between the feature vectors (Y1, Y2, Figure 17) of different indicator candidate products. If are closer than a certain level, it is determined that these indicator candidate products share external characteristics, and these different indicator candidate products can be selected as trend indicator products.

As described above, trend indicator products can be selected for each product category. For example, they are fashion items whose purchase value is determined by their appearance itself and whose class classification can be clearly determined from their external characteristics, such as earrings. It can also be applied to small accessories (see Figures 12a to 12c) or clothing such as children's clothing (Figure 18) or T-shirts where the flat design determines the external features. For example, several typical design elements excluding color are used. The probability of belonging to each class or the probability of belonging to each class output through a convolutional neural network model learned to recognize several typical designs such as plain patterns, striped patterns, check patterns, and frill decorations. By using a one-dimensional feature vector (Y1, Y2, Figure 17) whose probability is displayed at the index of each class, and comparing the Euclidean distance (S) and the reference value between different candidate index products, different candidates It is possible to determine whether external characteristics are shared between index products.

Referring to Figures 16 and 17, in one embodiment of the present invention, a group of products for which the buyer's preference changes rapidly are recognized as index candidate products, and products with common characteristics among this group of index candidate products are selected. It can be selected as a trend indicator product. Here, the group of products for which the buyer's preference changes rapidly includes the first group of indicator candidate products for which the buyer's preference rapidly increases, and the first group of index candidate products for which the buyer's preference rapidly decreases. It may include two groups of indicator candidate products, and among the first group of indicator candidate products, the first group of indicator candidate products that have common features, mainly in appearance, are considered trend indicator products of the first group. Separately from the indicator candidate products of the first group, the indicator candidate products of the second group that have common characteristics among the indicator candidate products of the second group can be selected as the trend indicator products of the second group. . At this time, the external characteristics of the index candidate products of the first and second groups are obtained by using the images of the index candidate products of the first and second groups as input, and using the convolution model of the server as a feature map or feature vector (Y1, Y2). , Figure 17) can be extracted, and the Euclidean distance (S, Figure 17) between different feature maps or feature vectors (Y1, Y2, Figure 17) can be calculated (similarity calculation). In this case, the indicator candidate of the first group The degree of similarity is calculated between products, and according to the results, the trend indicator products of the first group are selected. Separately from the trend indicator products of the first group, the degree of similarity is calculated between the index candidate products of the second group, and the trend indicator products of the first group are selected. Depending on the results, the second group of trend indicator products can be selected. As such, in one embodiment of the present invention, the calculation of similarity is performed only between the indicator candidate products of the first group and between the indicator candidate products of the second group, and between the indicator candidate products of the first group and the second group. Similarity calculation is not performed between index candidate products. The first group of trend indicator products are products for which buyers' preferences are rapidly increasing. Conversely, the second group of trend indicator products are products for which buyers' preferences are rapidly decreasing, so buyers should choose products of the corresponding trend. This is because we need to respond differently to each other.

<Points of one-to-one relationship between purchaser and purchaser>

Hereinafter, with reference to FIGS. 19A to 21B, points accumulated, deducted, and managed in a one-to-one relationship between a purchaser and a purchaser will be described.

Figures 19A to 19C show diagrams to explain points accumulated, deducted, and managed in a one-to-one relationship between a purchaser and a purchaser.

Figure 20 shows a diagram to explain the structure of point data stored in the server and the registration of points related to offline purchasing activities.

Figures 21a and 21b show diagrams showing an example in which point details of the purchaser and the purchaser are displayed on the web of the purchaser and the purchaser according to the account type logged in to the server.

Referring to FIGS. 19A to 21B, in one embodiment of the present invention, the server is provided individually from each purchaser and can manage point details on a per purchaser basis that are individually accumulated and deducted on a per purchaser basis. For example, in one embodiment of the present invention, the server may manage the details of points accumulated and deducted in a one-to-one relationship (FIGS. 19A to 19C) between the buyer and the purchaser who registered the purchase product. For example, the points The details can be stored in connection with the buyer's account information and the purchaser's account information related to each individual purchase activity, and the point details can be understood as the accumulated total of point information accumulated or deducted for each individual purchase activity. there is. For example, in one embodiment of the present invention, point details may mean time-series accumulation of point information accumulated or deducted from one purchase activity.

In one embodiment of the present invention, the server may provide a commerce relay service for one commerce zone, and the target for which the server relays commerce may be limited to one commerce zone, and accordingly, on the server The number of purchasers registering products may be limited. If, unlike in the present invention, a commerce relay service is not provided to buyers who are geographically tied to one commerce zone, but a commerce relay service is provided to purchaser members registered on the server without regional boundaries, , managing point details on a one-to-one basis for a large number of purchaser members with small sales volume wastes the memory resources of the server and reduces the efficiency of point details management, so in one embodiment of the present invention, one regional While providing a relay service for the commercial area, point details can be managed on a one-to-one basis (Figures 19a to 19c) for a relatively small number of purchaser members who have offline stores in the commercial area, as described later. , point details are managed by accumulating/deducting individual points generated from online and offline transactions of each purchasing member, so the sales volume of each purchasing member can be maintained above an appropriate level for point activation, and point In that the use of is activated, the efficiency of managing point details can be improved.

As shown in FIG. 20, in one embodiment of the present invention, each purchaser that registers a product on the server can operate an offline store within one commercial zone, and the server uses the products registered on the server. Along with points acquired through online purchases, points acquired through offline purchases at offline stores without going through the server can be accumulated and managed in an integrated form as point details, and accumulated points can also be stored on products registered on the server. It can be used for online purchases and offline purchases at offline stores, and the accumulated points can be deducted from the payment amount. When making an online purchase, the above server can check the information of the buyer who purchased the product and the information of the purchaser who registered the product. Accordingly, the details of points accumulated and deducted through each purchase activity are stored in a one-to-one relationship between each buyer and the purchaser. It can be managed in (Figures 19a to 19c). However, since the above server cannot confirm purchase activities that occur at each purchaser's offline store, the accumulation and deduction of individual points through the corresponding purchase activity is performed on the server according to the account information of the purchaser (purchase member) logged in to the server. can be registered, and the accumulation and deduction of points registered by the purchaser itself is notified to the purchaser member for whom the accumulation and deduction of points is made, or, for example, by the server that has searched the registration information of the purchaser member. The point details of the relevant purchaser are notified to the relevant purchaser member through the member's terminal, for example, a computing device including a mobile device such as a mobile phone, or according to the account information of the purchaser (purchaser member) logged in to the server at a later date. It can be displayed as .

With the above server, even if the purchaser member registers the accumulation and deduction of points for buyers who individually participated in the purchasing activity due to purchase activities that occur at the offline store of the purchaser (purchaser member), direct manual work by such purchaser member is not possible. Errors in the registration of accumulation and deduction of points cannot be confirmed, and since it may be inconvenient for each purchasing member to directly register accumulation and deduction of points, direct manual processing is done through the transaction receipt issued for the relevant purchase activity. can be avoided, and errors in registration caused by direct manual work can be prevented. For example, the purchaser member scans the receipt issued due to purchase activities at his/her offline store, for example, by scanning the code (1-dimensional barcode or 2-dimensional QR code) recorded on the receipt. This way, direct manual work can be avoided.

Referring to FIG. 20, in one embodiment of the present invention, in the server, for each individual point associated with each purchase activity, a series of items that participated in each purchase activity, i.e., product registration of the purchaser and purchase of the registered product, are stored in the server. A one-to-one relationship between a purchaser and a buyer related to purchasing activities can be stored. For example, the point details stored in the server are the total point data generated for each individual point by linking the purchaser's account information and the purchaser's account information related to one purchase activity in which point accumulation or deduction occurred. This may be applicable, and in the point data stored for each individual point, the account information of the purchaser who paid the corresponding points and the account information of the purchaser who accumulated the corresponding points are stored in different indexes (purchase index and buyer index). It may be. Referring to Figure 21a, the member's point details may be displayed on the member's terminal or the web according to the account information logged into the server, and the server searches the purchaser index of point data according to the account type and While displaying the point details linked to the purchaser as a whole, it is possible to search the buyer index of the corresponding point data and display the point details separately for each buyer. Similarly, referring to FIG. 20b, the server may display the member's point details on the member's terminal or the web according to the account information logged into the server, and the server may display the point data purchaser according to the account type. By searching the index, you can display the point details linked to the relevant purchaser as a whole, and by searching the purchaser index of the corresponding point data, you can display the point details separately for each purchaser. At this time, the point details provided to the purchaser member may refer to the total of individual points accumulated or deducted due to the purchasing activity of the purchaser member, but the accumulated or deducted points for all purchasers should be displayed as one item. Rather, for example, accumulated or deducted points can be displayed individually for each purchaser, for example, points accumulated or deducted for different purchasers can be displayed as separate items, and the buyer can By individually displaying the points that can be used for each purchaser for a member, it can be provided as reference material for the purchaser's future purchase activities.

In other words, referring to FIG. 21A, the server searches the purchaser index of point data according to the account type logged in to the server, and provides point details based on a set of point data stored in connection with the corresponding purchaser. Although it is displayed collectively, the point details can be displayed separately for each buyer based on the buyer index of each group of point data. Similarly, referring to Figure 21b, the server searches the buyer index of point data according to the account type logged in to the server, and provides point details based on a set of point data stored in connection with the corresponding buyer. It is displayed overall, but the point details can be displayed separately for each purchaser based on the purchaser index of each of the above group of point data.

In one embodiment of the present invention, the server manages point details in a one-to-one relationship between individual purchasers and individual buyers (FIGS. 19A to 19C), meaning that the points accumulated for a purchaser related to one purchase activity are the corresponding This means that it can be used for future purchase activities from the purchaser, and that discount benefits through deduction of points are provided only during future purchase activities from the relevant purchaser. Accordingly, as a purchaser, points that can be deducted from future purchase activities are provided. It may be necessary to check the accumulated purchaser. In one embodiment of the present invention, managing point details in a one-to-one relationship between individual buyers and individual purchasers means providing general-purpose points for online purchase activities on the server and deducting points paid from future online purchase activities. There is a difference from what is possible, and in one embodiment of the present invention, individual points accumulated or deducted from online purchases and offline purchases while managing point details in a one-to-one relationship between each individual buyer and each individual purchaser (FIGS. 19A to 19C). In online and offline purchases that are managed comprehensively and only target purchasers who have paid the corresponding individual points for each individual point, discount benefits may be provided through point deduction based on the accumulated points for the corresponding purchaser. will be.

In one embodiment of the present invention, the server can output point details for purchaser members according to the type of account information logged into the server, for example, points where individual point information forming the point details is recorded. Point details are calculated by referring to the data, and the calculated point details can be displayed on the purchaser member's terminal or on the web. At this time, the server may display the point details on the relevant purchaser member's terminal or on the web, and calculate the point details for each purchaser and output individual point details.

Referring to Figure 21b, in one embodiment of the present invention, each point details are displayed for each purchaser where purchase activities have been made to date, and the server displays each purchaser regardless of the location information of each purchaser. The accumulated point details can be displayed, or, as shown in FIG. 7, the accumulated point details (0P, 100P, 1000P) can be displayed for each purchaser along with the location information of each purchaser. For example, the server may provide a street view of a commercial area that provides a commercial transaction relay service, and match and display location information of purchasers and accumulated point details for each purchaser on the street view. For example, the server displays the street view of the commercial district providing relay services and the location information of each purchaser within the street view, allowing a virtual shopping experience even without directly visiting the commercial district relayed by the server. For example, on the street view, a hyperlink to a web document containing product information (link information) of each purchaser is displayed, and a click event (onclick or mouseover) for the corresponding hyperlink is displayed. Accordingly, a web displaying product information of the relevant purchaser can be displayed on the member's terminal.

Referring to FIG. 7, in one embodiment of the present invention, the location information of the purchaser displayed on the street view, along with the street view of the commercial transaction area relayed by the server, includes a specific area set on the image of the purchaser or the location of the purchaser. Product information of the relevant purchaser according to a click event occurring on a web element displayed on the image (a text hyperlink such as the purchaser's name or an image hyperlink such as the purchaser's trademark containing a symbol or figure symbolizing the purchaser) By displaying (image hyperlinks to selectable product images), product selection can be made through a second selection path in which product selection is initiated from the location information of each purchaser (street view showing the location of the purchaser). . For reference, in one embodiment of the present invention, according to the purchaser member's selection, product selection is made from information on the product itself, regardless of the location information of the purchaser, for example, rather than through a street view on which the location information of the purchaser is displayed. A second selection path (see FIG. 7) in which product selection is initiated from the first selection path (see FIG. 6) or alternatively, from a street view displaying the location information of the purchaser, for example, the location information of the purchaser. By this, product selection can be made, and on the street view displayed on the buyer's terminal or web in the second selection path (see FIG. 7), the point details for each purchaser (0P) along with the location information of each purchaser , 100P, 1000P), the accumulation of points that can be deducted from future purchase activities can be displayed together, and as a buyer, the accumulated points are relatively high, so by selecting the purchaser where the purchase activity is concentrated, priority is given to future purchases. Activities can be taken into consideration, and discount benefits can be provided through deduction of accumulated points.

For example, in one embodiment of the present invention, the server refers to a web document stored on the server and provides multimedia content such as an image or video that provides a street view as a lower layer according to priority such as z-index. It can be provided as a web element that overlaps with the street view as an upper layer and is displayed in a specific area of the street view or overlapping on the street view, such as a text hyperlink such as the company's name or an image hyperlink such as the buyer's trademark. can be displayed, and in addition, the accumulated and deductible point details (0P, 100P, 1000P) for each purchaser can be displayed on the location of the purchaser that is displayed overlapping on the street view, and information about these points can be displayed. The display is point history information, that is, a hyperlink (e.g., a hyperlink connected to the detailed information of point details (0P, 100P, 1000P) stored in a one-to-one relationship between the purchaser and the purchaser (Figures 19a to 19c) (e.g., in the total amount of points) It can be displayed as a text hyperlink, 0P, 100P, 1000P), similar to the string about the point, and detailed information of the point details (0P, 100P, 1000P) is calculated from the point data according to the click event that occurred in the hyperlink about the point. Detailed information of point details (0P, 100P, 1000P) can be displayed on the web. For example, in one embodiment of the present invention, the server provides a text hyperlink (0P, 100P) linked to the detailed information of the point details (0P, 100P, 1000P) along with the location information of the purchaser on the street view of the commercial district. , 1000P) can be displayed, and the text hyperlink (0P, 100P, 1000P) can be displayed as a string indicating the total accumulated points for the purchaser for which location information is displayed.

Referring to FIG. 9, in one embodiment of the present invention, the server can display the location of the purchaser for each theme product along with the overall view of the commerce area where commerce is relayed. For example, the server divides the window and displays the upper part. The frame provides an overall view of the commercial area, and the lower frame can display a web marked with radio buttons from which various themes can be selected on the buyer's terminal. At this time, according to the selection of different theme products that are selectively displayed with radio buttons, the server provides multi-media content such as images that provide an overall view as a lower layer according to priority such as z-index. By implementing a web document linked to the selected radio button as the upper layer, the upper layer containing web elements (image hyperlinks such as dot marks) that display the location of the purchaser handling the selected theme product can be viewed as an overall view of the lower layer. Arrange them so that they overlap, and according to the click event generated from the image hyperlink of the upper layer, the purchaser's product (selected theme product) corresponding to the location of the purchaser displayed in the overall view is displayed on the tab (window) where the overall view is displayed. It can be displayed overlapping or on a new tab (window) that is different from the tab (window) where the full view is displayed. At this time, the upper layer, which is displayed overlapping the overall view displayed as the lower layer, contains an image hyperlink that displays the location of the selected purchaser, as well as a string related to the total number of accumulated and deductible points for each applicable purchaser. Text hyperlinks (100P, 500P, 1000P) can be displayed, and a one-to-one relationship between the corresponding purchaser and buyer according to the click event generated from the text hyperlink (100P, 500P, 1000P) related to the point (FIGS. 19A to 19) As detailed information on the point details (100P, 500P, 1000P) is calculated from the point data stored in 19c), the detailed information on the point details (100P, 500P, 1000P) can be displayed on the buyer's web.

The present invention has been described with reference to the embodiments shown in the accompanying drawings, but these are merely illustrative, and various modifications and equivalent embodiments can be made by those skilled in the art. You will understand the point.

U: Delivery unit C: Delivery center
U1: Primary delivery unit C1: Primary delivery center
U2: Secondary delivery unit C2: Secondary delivery center

Claims (12)

A relay service system that includes a server that provides a commercial relay service between multiple purchasers and multiple buyers intensively located within one commercial area,
Perform sequential clustering for short-distance delivery by collecting different delivery target products from different purchasers and concentrating them into one delivery location.
For the forwarding target products or the forwarding target products excluded from the previous stage of clustering, the product with the largest volume is set as the forwarding center, and the volume limit is increased by advancing from the near to the far distance according to the distance from the forwarding center. Sequential clustering is performed to group each product into the same delivery unit until it is filled.
The server sequentially sets each delivery center and each delivery unit with each delivery center as the center,
For the products subject to delivery, the largest product is set as the primary delivery center, and a primary delivery unit with the primary delivery center as the center is prioritized,
Excluding the above-mentioned primary delivery unit, the product with the largest volume among other delivery target products is set as the secondary delivery center, and a secondary delivery unit is set with the secondary delivery center as the center,
Characterized by setting the 3rd delivery unit with the 3rd delivery center being the product with the largest volume among other delivery target products excluding the 1st delivery unit and the 2nd delivery unit, and the 3rd delivery center as the center. , relay service system. delete According to paragraph 1,
The server is characterized in that it requests distribution of transportation resources suitable for long-distance transportation of small-volume lightweight products, as the lower-priority delivery unit established in the latter half of sequential clustering includes lightweight products scattered over a wide area. Relay service system. According to paragraph 1,
The relay service system is characterized in that the server does not group specific products exceeding a limited distance from the delivery center within the limit volume into the same delivery unit. According to paragraph 4,
The server is a relay service system characterized in that the specific product is grouped together into a delivery unit to which other products that are closest to each other belong. According to clause 5,
A relay service system, characterized in that the delivery unit to which the other nearest product belongs includes the specific product and exceeds a limit volume. According to clause 6,
A relay service system, characterized in that the server requests the distribution of transportation resources suitable for short-distance transportation of bulky and heavy products to the delivery unit exceeding the limit volume. According to paragraph 4,
The server calculates the difference between the volume of the delivery unit that excludes a specific product exceeding the limit distance and the limit volume,
Depending on the difference between the volume of the forwarding unit and the limit volume, the forwarding unit is released for a forwarding unit whose volume is relatively small compared to the limit volume to the extent that it is judged to be inefficient to bundle the forwarding unit into one forwarding unit. A relay service system characterized in that the products of the released delivery unit are bundled into other nearest neighboring delivery units. According to paragraph 1,
The server groups specific products that exceed the limit distance from the delivery center within the limit volume into the same delivery unit,
A relay service system, characterized in that it requests distribution of transport resources suitable for long-distance transport for collection units exceeding the limit distance. According to paragraph 1,
As a specific product exceeding the limit distance from the delivery center within the limit volume is captured, the server releases the delivery unit set in the previous step,
A relay service system, characterized in that reclassification is performed to reset a delivery unit with the specific product as the center and the delivery center as the center. According to clause 10,
The server performs pre-processing clustering to sequentially set a delivery center based on volume and set a delivery unit with the delivery center as the center, while ensuring that the server is not grouped into a delivery unit beyond the limit distance from each delivery center. Recognize specific products that are not
Among the specific products, the product with the largest volume is set as the primary delivery center, and a primary delivery unit is set with the primary delivery center as the center,
Except for the primary delivery unit, the product with the largest volume among other specific products is set as the secondary delivery center, and a secondary delivery unit is set with the secondary delivery center as the center,
If there are no other specific products other than the 2nd delivery unit, the product with the largest volume among the other products excluding the 2nd delivery unit is the 3rd delivery center, and the 3rd delivery center is the 3rd delivery center. A relay service system characterized by setting a collection unit. According to clause 11,
The server performs pre-processing clustering to sequentially set a transmission center based on volume and set a transmission unit with the transmission center as the center, but is not grouped into a transmission unit by exceeding the distance limit from each transmission center. A relay service system characterized by performing pre-processing clustering until no other products exist except for a specific product.

Images