TWI684145B - A system for configuring storage spaces of warehouse - Google Patents

Abstract

The invention discloses a system for configuring storage spaces of warehouse, the system comprises a storage database, a social network internet post analysis module, a stock forecasting module, a warehouse data calculation module and a storage configuration module, wherein the social network internet post analysis module provides a product internet post from an external social network, the stock forecasting module provides a stock forecasting data, and the warehouse data calculation module provides a commodity data matrix about the state of the goods in the warehouse, and finally, the storage configuration module comprehensively considers the data provided by the storage database, the stock forecasting module, and the warehouse data calculation module to generate a warehouse storage location configuration information, thereby achieving the purpose of optimizing the storage location configuration.

Description

Warehouse storage configuration system

The invention relates to a storage allocation mechanism related to storage management, in particular to a storage allocation system that considers commodities, storage, and external information as storage allocation.

In the logistics industry, we usually face many problems of stocking and warehousing management, and in traditional warehousing management, we need to rely on manual experience to plan and adjust the warehousing configuration. Therefore, today's warehousing management has faced many challenges. If it can not be adjusted in real time with the fluctuation of product sales, there will be excess storage space in the off-season, but the problem of insufficient storage space will occur in the peak season. For example, the storage locations of goods with high shipping correlation are very far apart. Reduce the efficiency of picking. In addition, some products are similar in appearance and difficult to distinguish, which makes it easy to make mistakes when picking, or the price difference between adjacent goods is large, causing a large cost loss when picking wrongly. For example, items with high turnover rate are placed at the entrance and exit Farther away, it will reduce the picking efficiency, and the heavier goods will be placed above the lighter goods when picking, which will cause cargo damage. All of the above situations highlight the importance of warehousing and storage management. From the perspective of stocking, due to factors such as unknown future sales of new commodities, unknown correlation with other commodities, high utilization rate of storage space, insufficient storage space and other factors, it is difficult to decide the storage location of new commodities, and in the face of the aforementioned storage management Due to factors, it is difficult to make the goods properly stored in the warehouse, which will hinder the management of the storage allocation and management of incoming and outgoing goods.

According to this, in many situations such as the change in product sales and the entry of new products into the warehouse, the storage position of the warehoused goods needs to be adjusted accordingly, and the storage position adjustment will affect the subsequent picking and shipping operations, which will affect the overall shipping efficiency. In addition, many factors must be considered when adjusting storage locations, so it requires planning by experienced professionals. However, in today's society where all industries are facing human resource shortages, the logistics industry is also unavoidable to avoid manpower shortage. Therefore, how to analyze through data Information science methods, such as stock forecasting, take into account relevant factors, and assist logistics operators in planning storage space allocation, thereby achieving the goal of optimizing storage space allocation and improving the efficiency of picking and shipping.

The purpose of the present invention is to propose a storage and storage location allocation system, which uses basic information of commodities, previous historical data, sales information, and storage space data, etc., together with other external information to achieve the purpose of optimizing storage allocation.

In order to achieve the above or other objects, the present invention proposes a storage location configuration system, including: a storage database, a social network sound volume analysis module, a stock prediction module, a storage data calculation module, and a storage location planning module , Where the warehousing database is used to store basic commodity data, historical promotional data, historical order data, and warehousing space data, and the social network sound volume analysis module is used to obtain external social network article data, and According to the basic data of the commodity in the warehouse database, the correlation between the article data and the basic data of the commodity is analyzed to generate the sound volume data of the commodity; the stock prediction module is based on the basic data of the commodity and the history of the warehouse database Promotional data and the sound volume data of the commodity based on the social network sound volume analysis module to calculate and obtain stock forecast data; the storage data calculation module based on the historical order data and the basic data of the commodity in the warehouse database The commodity data matrix related to the status of the goods in the warehouse; and the storage location planning module based on the stock forecast data of the stock forecast module, the commodity data matrix of the storage data calculation module, and the commodity in the warehouse database The basic data and the storage space data are analyzed and calculated to generate warehouse storage location information.

In one embodiment, the social network sound volume analysis module further includes: a first webpage browsing unit for obtaining the article data; and for analyzing the association between the article data and the commodity basic data to obtain the commodity sound A semantic analysis unit for volume data; and a first memory unit including an article database for storing the article data and a commodity volume database for storing the commodity volume data.

In another embodiment, the stock forecasting module includes: a stock forecasting calculation unit and a second memory unit, wherein the stock forecasting calculation unit is used for the basic data of the commodity, the historical promotional data and the sound volume data of the commodity Input to a prediction model to calculate the stock prediction data, and the second memory unit is used to store the stock prediction data.

In another embodiment, the stock forecasting module further includes a second browsing webpage unit for acquiring weather data of an external website for the stock forecasting calculation unit to analyze and calculate the basic data of the product, the historical promotional data, and the product Sound volume data and the weather data to obtain the stock forecasting data.

In another embodiment, the storage location planning module further includes a planning unit that generates an appearance similarity matrix and a commodity price difference matrix, wherein the appearances classified by the text description in the basic data of the commodity are similar in appearance The sex matrix is generated by the planning unit based on the commodity basic data and the storage space data, and the commodity price difference matrix is the planning unit based on the normalized price difference between the commodity in each storage and its neighboring commodity Producer.

In another embodiment, the planning unit performs storage location cost calculation based on the stock forecast data, the product data matrix, the appearance similarity matrix, and the product price difference matrix, so as to obtain the warehouse storage location information, The storage allocation cost calculation is an algorithm to calculate and calculate the cost value. The cost value is sent back to the appearance similarity matrix and the commodity price difference matrix as the basis for consideration. After multiple iterations, the cost with the lowest cost value is generated. Warehouse location configuration information.

The storage and storage allocation system proposed by the present invention optimizes storage allocation through the integration of multiple messages, so that it can achieve a reduction in stocks during the off-season and adjust the corresponding stocks of goods in response to weather or festivals. In addition, it can avoid shipping When the products with high correlation are stored far apart, the products with similar appearance and indistinguishable are closely spaced, the price difference between adjacent products is large, and the turnover rate is high. It is placed on the lighter goods and other inappropriate storage methods, so that the merchant can achieve the purpose of effective management of the goods or goods.

The technical content of the present invention will be described below with specific specific implementation forms, and those skilled in the art can easily understand the advantages and effects of the present invention from the content disclosed in this specification. However, the present invention can also be implemented or applied by other specific embodiments.

The modules, units, devices, etc. of the present invention include a microprocessor and a memory, and algorithms, data, programs, etc. are stored in a memory or chip, and the microprocessor can load data or algorithms or programs from the memory to perform Processing such as data analysis or calculation will not be repeated here. For example, the storage database, the social network sound volume analysis module, the stock prediction module, the storage data calculation module, and the storage location planning module of the present invention include a microprocessor and a memory, etc., and each module Each unit of the unit executes the analysis operation in this way, so the hardware details of the unit or module described in the present invention can also be implemented in the same way.

Figure 1 illustrates the warehouse storage system of the present invention. As shown in the figure, the warehouse storage location configuration system 100 includes a warehouse database 1, a social network sound volume analysis module 2, a stock prediction module 3, a warehouse data calculation module 4, and a storage location planning module 5. The warehousing and storage location configuration system 100 can be integrated based on the commodities and historical data in the warehouse and external information to obtain better product configuration results.

The warehousing database 1 is used to store basic merchandise data 11, historical promotional data 12, historical order data 13 and storage space data 14. Among them, the basic merchandise data 11 can be used for comparison between commodities, and the historical promotional data 12 can be used as stock considerations. The historical order data 13 can provide the basis for purchase, and the storage space data 14 provides the existing storage configuration information. In one embodiment, the basic commodity data 11, historical promotion data 12, historical order data 13 and storage space data 14 can be stored in the memory, when the aforementioned data is to be used, they can be loaded from the memory to (load) to microprocessor. The following modules, units or devices are processed in the same or similar manner.

The social network sound volume analysis module 2 is used to analyze the article data obtained from the external social network and the commodity sound volume data of the commodity basic data 11 obtained from the warehouse database 1. In short, the basic product information 11 can get information about the types of goods in the store, and how these products are related to the purchase volume and storage configuration in the consumer evaluation, so the social network sound volume analysis module 2 will go to the external community Networks, such as forums or discussion forums, obtain article information about products, and then obtain product sound volume data, that is, reviews about the product, as a consideration for future product purchases and configurations. Specifically, before the analysis, the common adjectives, adverbs and other words will be marked with emotional scores in advance to form a commonly used database, and then the social network voice analysis module 2 will obtain the articles of the group network through automatic web browsing Data, you can know the product sound volume data by comparing the text in the article data with the information stored in the common database. Regarding the more detailed calculation method of sound volume data of commodities, more detailed examples will be explained later in the present invention. In one embodiment, analysis of article data obtained from an external social network and commodity basic data 11 obtained from the warehousing database 1 can be stored in the memory, and when the aforementioned data are desired, they can be used separately Load from memory to the microprocessor. The following modules, units or devices are processed in the same or similar manner.

The stock prediction module 3 receives the basic commodity data 11 and historical promotion data 12 of the warehousing database 1 and the commodity volume data from the social network sound volume analysis module 2, and obtains stock prediction data through analysis and calculation, where, The stock forecasting data refers to the reference basis for the store’s future stocking, so in addition to the product sound data related to consumer opinions, of course, it will also consider the historical promotional materials related to the product. In simple terms, the basic product data 11 and historical sales data 12 A prediction model can be obtained through a recurrent neural network method, and then when the quantity of goods to be stocked is to be predicted, the information of the goods to be purchased is substituted into the prediction model to obtain stock prediction data. Under this mechanism, this can be obtained In what period of sales, the purpose of determining the amount of purchases. In one embodiment, the sales forecast data, the quantity of goods stocked, and other sales conditions can be stored in the memory, and when the aforementioned data is to be used, they can be loaded from the memory to the microprocessor, wherein the forecast model of the stock 3 Executed in the microprocessor. The following modules, units or devices are processed in the same or similar manner.

The warehousing data calculation module 4 parses the historical order data 13 and basic commodity data 11 of the warehousing database 1 to obtain a commodity data matrix related to the status of the goods in the warehouse. In detail, the analysis of historical order data 13 and basic product data 11 can obtain the order status of each product. This is related to the storage of goods. For example, infrequent order indicates that the sales speed is relatively slow, that is, the shipping probability is relatively Low, so put the goods in storage more internally to avoid affecting the shipment of popular goods. In one embodiment, the warehouse data calculation module 4 is executed in the microprocessor. The following modules, units or devices are processed in the same or similar manner.

The storage location planning module 5 receives and uses an algorithm to analyze and calculate the stock prediction data of the stock prediction module 3, the commodity data matrix of the storage data calculation module 4, and the basic commodity data 11 and The storage space data 14 is used to generate warehouse storage location information. The storage location planning module 5 is used to plan the storage location configuration of the warehouse. It obtains the stock forecast data, the commodity data matrix, the basic commodity data 11 and the storage space data 14. Through the algorithm, the final warehouse storage location information is generated. That is to say, the basic product data 11 and the storage space data 14 can obtain the similarity in appearance and the price difference of the products. The above information will affect the storage location of the goods. The data matrix can produce a combination of various storage configurations. Taking into account the above appearance similarity and commodity price differences, the cost value of this configuration can be calculated through the formula. After several iterations, the storage with the lowest cost value can be calculated. The bit configuration becomes the most suitable storage configuration. Among them, heuristic algorithms such as genetic algorithm and particle swarm algorithm can be used for the algorithm. In one embodiment, the storage location planning module 5 is executed in the microprocessor. The following modules, units or devices are processed in the same or similar manner.

It can be seen from the above that through integration of internal and external information in the warehouse, in addition to considering the consumer's mentality or evaluation, it also considers the order quantity and sales of the goods and the current configuration results of the warehouse, and then finds the best warehouse storage configuration information through the algorithm , So that the store can have a reference for subsequent ordering and storage of goods.

FIG. 2 is a structural diagram illustrating a specific embodiment of the storage location configuration system of the present invention. As shown in the figure, the warehousing database 1 stores basic commodity data 11, historical promotion data 12, historical order data 13, and storage space data 14. Further, the basic product information 11 can include product number, product name, weight, selling price and text description and other distinguishable product information. The basic product information 11 can be used as a basis for querying other information, such as to an online chat room Or get product reviews in the forum, or get sales, promotions, and inventory status based on the product name.

The historical promotional data 12 can include the product number, order date, order number, and order quantity. The historical promotional data 12 can be used as a reference for the storage of the product. For example, when the sales period of a product is approaching, then when the product is purchased, the placement position must of course be The convenience of recent shipments is considered.

In addition, the historical order data 13 records the previous sales of the goods, which can include information such as product number, promotion date and sales price, and the storage space data 14 contains the storage number, coordinates and capacity, which are related to the location of the goods in the warehouse. And other information.

In addition to the aforementioned warehouse database 1, the warehouse storage location configuration system 100 also includes a social network sound volume analysis module 2, a stock prediction module 3, a warehouse data calculation module 4 and a storage location configuration module 5, wherein, The social network sound volume analysis module 2 includes a first web browsing unit 22, a semantic analysis unit 21, and a first memory unit 23. For example, in specific implementation, the semantic analysis unit 21 may be a microprocessor and a first memory The unit 23 may be a memory, and so on, which will not be described in detail below. The stock prediction module 3 includes a stock prediction arithmetic unit 32 and a second memory unit 33; and the storage location planning module 5 includes a planning unit 52. Subsequent descriptions will be given for each module of the warehouse storage location configuration system 100 in FIG. 2 in conjunction with FIGS. 3-6.

Please refer to Figures 2 and 3 together. Figure 3 is a structural diagram of the social network sound volume analysis module of the present invention. As shown in the figure, the social network sound volume analysis module 2 is used to obtain external social networks (for example: PTT, Dcard, Mobile01, Facebook, etc.) articles that mention basic product information (such as product name) Data, in addition, the social network sound volume analysis module 2 will also go to the warehousing database 1 to obtain the commodity basic data 11, thereby analyzing the article data and the commodity basic data 11 to generate commodity volume data.

The social network sound volume analysis module 2 may include a first browsing webpage unit 22 for searching each social network and obtaining article data, and a semantic analysis unit for analyzing article data and basic product data to obtain product sound volume data 21. and a first memory unit 23 including an article database 231 for storing the article data and a commodity volume database 232 for storing the commodity volume data. With regard to how the social network sound volume analysis module 2 executes the analysis process, the following will illustrate with examples.

The social network sound volume analysis module 2 will pre-create a common vocabulary, that is, pre-stored common adjectives and adverbs marked with emotional scores in the vocabulary. The vocabulary can also be used to add new adjectives and adverbs in the future. Among them, the labeling of emotion scores such as "good" is an adjective of emotion score "+5", "no" is an adjective of emotion score "-1", "very" is an adverb of emotion score "+2", etc., accordingly, When analyzing the article, use the semantic analysis module 2 to extract the sentences or paragraphs mentioned in the article based on the basic product information 11, and check the common adjectives in the sentence and the adverbs modified corresponding to the common adjectives to obtain the article information Emotional scores, which are then used to evaluate commodities to generate commodity sound volume data.

式1 In an embodiment, the social network sound volume analysis module 2 collects article data related to commodities in various social networks through the first web browsing unit 22, and stores the article data in the article database 231, the first memory unit 23 may include a pre-established lexicon, and the semantic analysis unit 21 analyzes the article data 11 from the warehouse database 1 and article data 231 from the article database 231, and then analyzes the adjectives and adverbs of the commodities in the article data. Emotion score, and the emotion scores of adjectives and adverbs are substituted into the emotion score calculation formula, and then the product sound volume data is generated and stored in the product sound volume database 232, the calculation formula is as follows: 1:

Formula 1

表示句子中第

個形容詞的情緒評分，

表示第

個形容詞前的副詞之情緒評分。 among them,

Indicates the first in the sentence

Sentiment scores of adjectives,

Means first

The emotional score of the adverb before each adjective.

Please refer to FIG. 2 and FIG. 4 together. FIG. 4 is a structural diagram of the stock prediction module 3 of the present invention. The stock prediction module 3 receives the basic commodity data 11 and historical promotion data 12 from the warehouse database 1 and the commodity volume data from the social network sound volume analysis module 2, and calculates to obtain the stock prediction data 331, based on which The stock forecasting module 3 obtains stock forecasting data 331 by calculating and analyzing the above data to predict the required stock quantity of the goods. In one embodiment, the stock forecasting module 3 includes a stock forecasting calculation unit 32 and a second memory unit 33, wherein the stock forecasting calculation unit 32 is used to input basic product data 11, historical sales data 12 and product sound volume data To a prediction model, to calculate the stock forecasting data, that is, the stock forecasting calculation unit 32 uses the basic product data 11 and the historical sales data 12 to obtain the prediction model through a recurrent neural network method, and then will predict the purchase of product information It is substituted into the prediction model to obtain the stock prediction data, and the second memory unit 33 can be used to store the stock prediction data 331. In specific implementation, the second memory unit 33 can be a server database.

In another embodiment, the stock forecasting module further includes a second browsing webpage unit 34, which can be used to obtain weather data of an external website, where the weather data can include historical weather data 332 and weather forecast data 333, the stock prediction operation unit 32 can analyze and calculate the weather data, the commodity basic information 11, the historical promotion data 12, and the product sound volume data. Similarly, the stock prediction operation unit 32 takes into account the basic product information 11 and In addition to the historical promotional data 12, information such as historical weather data 332 and weather forecast data 333 is also added. The prediction model is also generated through a recurrent neural network, and the prediction model can enter the stock information of the expected purchase to obtain the stock. Forecast data 331. To put it simply, due to the influence of different weather, the stock of goods also needs to be adjusted. For example, rainy days require umbrellas and sunny days should strengthen sun protection. Therefore, the second browsing webpage unit 34 crawling weather data may include historical weather data 332 or weather forecast data 333 rainfall, maximum temperature or minimum temperature, etc., to estimate the quantity of goods in stock.

In addition, the stock prediction module 3 also includes a timer 35 that provides timing data, wherein the stock prediction operation unit 32 can retrieve the timing data as a reference basis for calculating the stock prediction data 331. In short, when the stock forecasting calculation unit 32 analyzes the timing data, it is based on the different demand for commodities on weekdays, holidays or holidays. Therefore, the stock forecasting calculation unit 32 analyzes the timing data to determine whether the stocking period is close to holidays or festivals , And then serve as a reference for predicting stocking, that is, when approaching the holiday, increase the stocking of merchandise related to the holiday. Specifically, the timing data of the timer 35 may include factors such as the date, day of the week, or whether it is a holiday, and may be combined with historical weather data 332 and weather forecast data 333 for the stock forecasting calculation unit 32 to use the timing data for analysis. That is to say, when the stock forecasting calculation unit 32 uses the built prediction model for stock forecasting, it will add the consideration of the timing data of the timer 35 to obtain the final stock forecast data 331.

In another embodiment, a forecasting model will be created before the commodity inventory prediction is performed, and the inventory prediction arithmetic unit 32 may input the commodity basic data 11, historical promotional data 12 and commodity sound volume data of the commodity that is secondary to inventory to The forecast model calculates the stock forecast data 331. Specifically, the establishment of the prediction model is to use the previously obtained basic data of the commodity 11, historical promotional data 12 and commodity sound volume data to normalize, that is, to generate usable information based on the relevance of the above data and return it After the neural network training, the aforementioned prediction model is generated, and the recurrent neural network is a long-short-term memory (LSTM) neural network, a recurrent neural network (RNN) or a sequence-to-sequence (seq2seq) model, or is based on Other neural networks with similar functions are implemented.

Further, the stock forecasting module 3 of this system is a stock forecasting method based on recurrent neural network. The stock forecasting method includes two parts: prediction model generation and future period prediction. For example, in the generation of prediction models, Perform the normalization of the historical data of the specified product to generate normalized historical data, and then use the normalized historical data (such as historical order quantity, historical weather data, commodity selling price, etc.) to train the recurrent neural network to obtain the prediction model, and in In the forecast of the future time period, the data of the future time of the specified product will be first processed to generate the normalized future data, and then the normalized future data (such as future order quantity, future weather data, product selling price, etc.) will be substituted into the recursive The prediction model obtained by neural network training is used to carry out the stock forecast in the future period, and finally the predicted stock quantity of the goods to be purchased will be obtained.

Specifically, taking the time axis from T-2 to T+1 as an example, on the T-1th day, the stock prediction calculation unit 32 receives the order quantity of the product on the T-2 day and analyzes the data on the T-1 day, whether Holidays or festivals, product selling price (promotional price or recommended selling price), product social network sound volume, maximum temperature, minimum temperature, and date or week, to generate the predicted stock quantity of the T-1th day, and then get The order quantity of the goods on the T-1th day; on the Tth day, the stock forecasting calculation unit 32 receives the order quantity of the goods on the T-1th day and analyzes the data on the Tth day, whether it is a holiday or a festival, the selling price of the goods (promotional price or suggestion) Price), the volume of the product social network, the maximum temperature, the minimum temperature, and the date or week, to generate the predicted stock of the product on the Tth day, and then get the order quantity of the product on the Tth day, and so on to obtain the timeline The stock quantity on the last date.

Please refer to FIGS. 2 and 5 together. FIG. 5 is a structural diagram of the storage data calculation module of the present invention. The warehousing data calculation module 4 can obtain the historical order data 13 and the commodity basic data 11 from the warehousing database 1 to obtain the commodity data matrix about the status of the goods in the warehouse by analyzing the historical order data 13 and the commodity basic data 11. The commodity information matrix includes a commodity turnover rate matrix 41 related to a single commodity turnover rate, an inter-product shipment correlation matrix 42 related to inter-product shipment correlation, and a normalized commodity weight matrix 43 related to product weight.

Specifically, the historical order data 13, the basic commodity data 11 and the storage space data 14 of the warehousing database 1 can be used by the warehousing data calculation module 4 to calculate a single commodity turnover rate, thereby generating a single commodity turnover rate matrix 41, in which a single commodity The turnover rate calculation formula is f _n1 , which is the number of product orders/total number of orders.

的全0矩陣

，其中，

為商品品項總數，矩陣內第i行第j列的元素以

表示，

；(2)若商品i和商品j出現在同一張訂單，則將

值加1；(3)執行步驟(2)，累計所有訂單內的所有商品；(4)找出將

內所有的元素的最大值

，並將

內的所有元素皆除以

；以及(5)矩陣R即為貨品間出貨相關性矩陣。透過上述步驟，將可得到有關貨品間出貨相關性之貨品間出貨相關性矩陣42。 The inter-shipment shipment correlation matrix 42 in the warehousing data calculation module 4 represents the inter-shipment shipment correlation, where the calculation analysis formula for the inter-shipment shipment correlation is f _n2 , which includes the following steps: (1) Create a

All 0 matrix

,among them,

Is the total number of product items.

Means,

; (2) If commodity i and commodity j appear in the same order, then

Add 1 to the value; (3) Perform step (2) to accumulate all commodities in all orders; (4) Find out

The maximum value of all elements in

And

All elements in are divided by

; And (5) Matrix R is the correlation matrix between shipments. Through the above steps, the inter-shipment shipment correlation matrix 42 related to the inter-shipment shipment correlation can be obtained.

The normalized product weight matrix 43 in the warehousing data calculation module 4 is used to normalize the weight of the product. Among them, the calculation analysis formula for the normalized product weight is f _n3 , which includes the following steps: (1) The weight in the basic data of the product , You can get a Nx1 commodity weight matrix; (2) get the minimum and maximum values of all elements in the matrix from the matrix; (3) subtract the minimum value of all elements in the matrix, and then divide by the maximum in the matrix Value minus the minimum value, you can get the normalized product weight matrix. Through the above steps, the normalized product weight matrix 43 related to the product weight will be obtained.

Please refer to FIGS. 2 and 6 together. FIG. 6 is an architecture diagram of the storage location planning module of the present invention. As shown in Figure 2, the storage location planning module 5 can receive the stock forecasting data from the stock forecasting module 3, the commodity data matrix from the storage data calculation module 4, and the basic commodity data 11 from the storage database 1 and The storage space data 14 is analyzed and calculated by an algorithm to generate warehouse storage location information. According to this, the system of the present invention calculates the stock forecasting data, commodity data matrix, and commodity basic data 11 and storage space data through the storage location planning module 5 14 It can predict the required stock quantity and better storage location, so to avoid the problem of excess or insufficient storage space, not only the effective use of storage space, but also improve the picking efficiency.

As shown in FIG. 6, the storage location planning module 5 includes a planning unit 52, which can generate an appearance similarity matrix classified according to the text description of the basic product data 11 based on the basic product data 11 and the storage space data 14 521 and a commodity price difference matrix 522 based on the normalized price difference between the commodity in each storage and its neighbors.

The calculation analysis formula of the appearance similarity matrix 521 is f _n4 , which is analyzed according to the following steps: (1) The products are clustered and grouped according to the text description in the basic data of the product 11 to obtain the classification value of the appearance similarity of the product; (2) The algorithm generates a set of storage configurations; and (3) Within the storage configuration, calculate the difference in the similarity classification value of the commodities in each storage and the commodities in the adjacent storage, which is the storage configuration The appearance similarity matrix 521.

The calculation analysis formula based on the commodity price difference matrix 522 is f _n5 , which is analyzed according to the following steps: (1) According to the commodity's recommended selling price in the basic information of the product minus the minimum value of all the suggested selling prices, divided by the recommended selling price The difference between the maximum value and the minimum value of the recommended selling price to obtain the normalized price of each commodity; (2) the algorithm generates a set of storage locations; and (3) within the storage configuration, calculate each storage location The normalized price difference between the commodity and the commodity in its adjacent storage is the commodity price difference matrix 522 configured for the storage.

式2 With the appearance similarity matrix 521 and the commodity price difference matrix 522, the planning unit 52 can perform the storage allocation cost calculation based on the stock prediction data 331, the commodity information matrix, the appearance similarity matrix 521, and the commodity price difference matrix 522 , In order to obtain the final warehouse location allocation information 51. In one embodiment, the above storage allocation cost calculation is an algorithm to calculate the cost value. Various storage allocation combinations will be used to consider the appearance similarity and commodity price difference of commodities. The cost formula f _{n6 is used} to calculate various configurations The cost value, the cost value will be sent back to the appearance similarity matrix 521 and the commodity price difference matrix 522 as a basis for consideration, after multiple iterations to generate the lowest cost value of the warehouse storage location information 51, wherein the cost calculation is based on the cost The formula f _{n6 is} used for calculation, and the calculation result is analyzed by the algorithm to calculate the cost value. For example, the algorithm may be a genetic algorithm, particle swarm algorithm, or other heuristic algorithm. In addition, the aforementioned cost formula f _{n6 is} as follows 2 shows:

Formula 2

With reference to the foregoing description, the planning unit 52 can decide the stock quantity of the new product based on the stock prediction data 331, taking into account the sound volume of the product on the social network, and comprehensively consider the required stock quantity with reference to the weather data and timing data, and then achieve The purpose of reducing the stock quantity in the off-season and adjusting the corresponding stock quantity according to weather or festivals, etc. In addition, the planning unit 52 can also use the product information matrix (commodity turnover rate matrix 41, inter-product shipment correlation matrix 42 and regularized products) Weight matrix 43), appearance similarity matrix 521, and product price difference matrix 522 are combined to avoid the separation of products with high shipping correlations that are far apart, similar appearance, and indistinguishable products are close or adjacent For the purpose of problems such as large commodity price differences, and planning unit 52 can also base on commodity basic data 11, historical order data 13 and normalized commodity weight matrix 43 to avoid items with high turnover rates being placed far away from the entrance and When heavy goods are picked, they are placed above lighter goods.

In summary, the warehouse storage allocation system of the present invention optimizes the storage allocation through the integration of multiple messages, can predict the required stocking quantity, and provides optimized storage for the stored or stocked goods, so avoid The problem of excess or insufficient storage space can also improve the picking efficiency. Through the storage and storage allocation mechanism of the present invention, it will be able to provide better stock management and storage allocation for stores.

The above-mentioned embodiments only exemplarily illustrate the principle and efficacy of the present invention, and are not intended to limit the present invention. Anyone familiar with this skill can modify and change the above-mentioned embodiments without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the rights of the present invention should be as listed in the accompanying patent application scope.

1‧‧‧ Warehouse database 100‧‧‧ Warehouse storage configuration system 11‧‧‧Commodity basic information 12‧‧‧ Historical Promotion Information 13‧‧‧Historical order information 14‧‧‧Storage space information 2‧‧‧Community network sound volume analysis module 21‧‧‧Semantic Analysis Unit 22‧‧‧First webpage browsing unit 23‧‧‧ First memory unit 231‧‧‧ Article database 232‧‧‧Commercial Sound Volume Database 3‧‧‧Stock Prediction Module 32‧‧‧ Stock prediction arithmetic unit 33‧‧‧Second memory unit 331‧‧‧Stock forecast data 332‧‧‧ Historical Weather Information 333‧‧‧ Weather forecast data 34‧‧‧Second page browsing unit 35‧‧‧Timer 4‧‧‧Warehouse data calculation module 41‧‧‧Commodity turnover rate matrix 42‧‧‧ Shipping correlation matrix 43‧‧‧Regular commodity weight matrix 5‧‧‧Storage configuration planning module 51‧‧‧ Warehouse storage configuration information 52‧‧‧Planning unit 521‧‧‧Appearance similarity matrix 522‧‧‧Commodity price difference matrix

Figure 1 is a system architecture diagram of the storage and storage location configuration system of the present invention;

Figure 2 is a structural diagram of a specific embodiment of the storage and storage location configuration system of the present invention;

Figure 3 is a block diagram of the social network sound volume analysis module of the storage and storage location configuration system of the present invention;

FIG. 4 is a structural diagram of a stock forecasting module of the warehouse storage location configuration system of the present invention;

Figure 5 is a structural diagram of a storage data calculation module of the storage and storage location configuration system of the present invention; and

FIG. 6 is a structural diagram of a storage location planning module of the storage and storage location configuration system of the present invention.

1‧‧‧ Warehouse database

100‧‧‧ Warehouse storage configuration system

11‧‧‧Commodity basic information

12‧‧‧ Historical Promotion Information

13‧‧‧Historical order information

14‧‧‧Storage space information

2‧‧‧Community network sound volume analysis module

3‧‧‧Stock Prediction Module

4‧‧‧Warehouse data calculation module

5‧‧‧Storage configuration planning module

Claims (13)

A warehousing and storage location configuration system, including: warehousing database, used to store basic commodity data, historical promotional data, historical order data, and warehousing space data; social network sound volume analysis module, used to obtain external social media The article data of the group network, and based on the basic data of the commodity in the warehouse database, to analyze the correlation between the article data and the basic data of the commodity to generate the sound volume data of the commodity; the stock prediction module is based on the warehouse data The basic data of the commodity and the historical sales promotion data of the library and the sound volume data of the social network sound volume analysis module are calculated to obtain the stock prediction data; the storage data calculation module is based on the storage database. The historical order data and the basic data of the product are analyzed to obtain a product data matrix related to the status of the goods in the warehouse; and the storage location planning module is based on the stock forecasting data, the product data matrix, and the basic data of the product and the Storage space data, analysis and calculation to generate warehouse storage location information. The storage location allocation system as described in item 1 of the patent application scope, wherein the social network sound volume analysis module includes: a first webpage browsing unit, which is used to obtain article data of the social network; The semantic analysis unit is used to analyze the association between the article data and the commodity basic data to obtain the commodity sound volume data; and the first memory unit includes an article database for storing the article data and for storing the commodity Commodity volume database for volume data. The warehousing and storage location configuration system as described in item 1 of the scope of the patent application, wherein the stock forecasting module includes: a stock forecasting calculation unit, which is used for the basic data of the product, the historical promotional data and the sound volume data of the product Input to a prediction model to calculate the stock prediction data; and the second memory unit is used to store the stock prediction data. The storage location allocation system as described in item 3 of the patent application scope, in which the prediction model is normalized by using the existing data of the basic data of the commodity, the historical promotional data and the sound volume data of the commodity obtained previously, Generated after recurrent neural network training. The storage location allocation system as described in item 4 of the patent application scope, wherein the recurrent neural network is a long-short-term memory (LSTM) neural network, a recurrent neural network (RNN) or a sequence-to-sequence (seq2seq) model . The storage location allocation system as described in item 3 of the patent application scope, wherein the stock forecasting module includes a second browsing webpage unit, which is used to obtain weather data of an external website for analysis and calculation by the stock forecasting arithmetic unit The basic data of the commodity, the historical promotional data, the sound volume data of the commodity and the weather data to obtain the stock forecast data. The storage location allocation system as described in item 6 of the patent application scope, wherein the weather data includes historical weather data and weather forecast data. The warehousing and storage location configuration system as described in item 3 of the patent application scope, wherein the stock forecasting module includes a timer that provides timing data, and the stock forecasting calculation unit retrieves the timing data to calculate the stock forecast Reference basis of information. The warehousing and storage location allocation system as described in item 1 of the patent application scope, wherein the product information matrix includes a product turnover rate matrix related to a single product turnover rate, an inter-product shipment correlation matrix related to inter-product shipment correlation, and Normalized product weight matrix on product weight. The storage location allocation system as described in item 1 of the patent application scope, wherein the storage location planning module includes a planning unit, which is generated based on the commodity basic data and the warehouse space data. The commodity basic The textual description matrix in the data describes the classification similarity matrix and the commodity price difference matrix by the normalized price difference between the commodity in each storage and its adjacent commodity. The storage location allocation system as described in item 10 of the patent application scope, wherein the planning unit executes the storage allocation cost based on the stock forecasting data, the commodity data matrix, the appearance similarity matrix and the commodity price difference matrix Calculate to obtain the storage location information of the warehouse. The storage location allocation system as described in item 11 of the patent application scope, wherein the storage allocation cost calculation is an algorithm analysis to calculate the cost value, and the cost value is sent back to the appearance similarity matrix and the commodity price difference The matrix is used as the basis for consideration, and after multiple iterations, the storage location allocation information of the warehouse with the lowest cost value is generated. The storage location allocation system as described in item 12 of the patent application scope, wherein the algorithm is a genetic algorithm or particle swarm algorithm.

Images