KR20210137264A - Big data based on-line shop AI keyword recommending system and method - Google Patents

Abstract

Disclosed are an artificial intelligence (AI) keyword recommendation for online shopping malls based on big data, which increases the number of times of a seller's product appearing in buyers' searches, and a method thereof. According to one embodiment of the present invention, a system using big data to recommend keywords of online shopping malls on the basis of AI comprises: a big data collection module collecting big data about seller products sold on an open market platform; a big data analysis module analyzing the big data to perform keyword analysis on the seller product; and a customer channel module recommending an exposure optimization keyword for the seller product on the basis of a keyword analysis result. The big data analysis module collects tag keywords matched for the seller product as a reference meta tag, and analyzes posts or images of influencers found by using the reference meta tag to extract influencer keywords, thereby performing keyword analysis.

Description

Big data based on-line shop AI keyword recommending system and method

The present invention relates to an online shopping mall operating system, and more particularly, to a system and method for recommending AI (artificial intelligence) keywords for an online shopping mall based on big data.

Recently, in the e-commerce market, from consumer-oriented simple search product purchases such as image search, product tagging, and application of big data product recommendation based on personal search history to the existing strong search base and price comparison level, artificial intelligence is provided to consumers through the AI open market. It is changing to a pattern that automatically recommends customized products.

AI open market technology is analyzed based on big data of various search histories such as search and purchase in the pre-consumption stage of the user. In this personalized AI open market market, it has become more important to manage sales products to be optimized for the AI open market than to expose push-type products through high-cost advertisement execution.

In the search-based AI open market, sales should be carried out through AI product optimization preferred by the AI system to be exposed to product recommendations and searches. Therefore, there is a need for a solution that exposes products to potential customers at a low cost and predicts sales of exposed products for small business sellers who are in the early stage of starting a business or burdened with advertising costs.

Korean Patent Laid-Open Patent No. 10-2018-0002944 (published on Jan. 09, 2018) - Big data recommendation system and recommendation method for optimal product suggestion for each individual preference

The present invention is a big data-based online big data-based online method that can increase the number of times a seller's product is exposed to a buyer's search by automatically recommending keywords optimized for search exposure to the seller through big data analysis in the open market. It is to provide a shopping mall AI keyword recommendation system and method.

Objects other than the present invention will be easily understood through the following description.

According to one aspect of the present invention, there is provided a system for recommending keywords according to artificial intelligence (AI) in an online shopping mall using big data, comprising: a big data collection module for collecting big data about seller products sold on an open market platform; a big data analysis module that analyzes the big data and performs keyword analysis on the seller product; and a customer channel module that recommends an exposure optimization keyword for the seller product based on the keyword analysis result, wherein the big data analysis module collects tag keywords matched for the seller product as a reference meta tag, and A big data-based online shopping AI keyword recommendation system is provided, which analyzes the posts or videos of the found influencers using the reference meta tag, derives them as influencer keywords, and performs keyword analysis.

One or more of power bloggers found on portal sites, SNS users who have more than a certain number of subscribers or followers on social network services, and video uploaders with more than a certain number of subscriptions found on video sharing sites can be found as influencers.

The more recent the registration date of the post or video of the influencer, the higher the number of views of the post or video, the higher the weight is applied to the word derived from the post or video to be derived as the influencer keyword. can

Words included in the text displayed in the video can be extracted from the video through image analysis, or words included in the description voice of the video uploader included in the video can be derived as the keywords of the influencer through voice analysis. .

Other aspects, features and advantages other than those described above will become apparent from the following drawings, claims, and detailed description of the invention.

According to an embodiment of the present invention, the number of times the seller's product is exposed to the buyer's search can be increased by automatically recommending keywords optimized for search exposure to the seller through big data analysis in the open market based on artificial intelligence (AI). has the effect of

1 is a block diagram of a big data-based online shopping mall AI keyword recommendation system according to an embodiment of the present invention;
2 is a block diagram of a big data collection module;
3 is a block diagram of a part for distributed storage of data in a big data collection module;
4 is a diagram showing a sharding structure for data distributed storage;
5 is a diagram showing a data distribution storage structure;
6 is a flowchart of a big data-based online shopping mall AI keyword recommendation method according to an embodiment of the present invention;
7 is a table showing a transaction database and frequent item sets for explaining a keyword recommendation process;
8 to 10 are views showing a seller product page, high-ranked search product information, and image recognition result screens, respectively.

Since the present invention can have various changes and can have various embodiments, specific embodiments are illustrated in the drawings and described in detail. However, this is not intended to limit the present invention to a specific embodiment, it should be understood to include all modifications, equivalents and substitutes included in the spirit and scope of the present invention.

When a component is referred to as being “connected” or “connected” to another component, it is understood that the other component may be directly connected or connected to the other component, but other components may exist in between. it should be On the other hand, when it is mentioned that a certain element is "directly connected" or "directly connected" to another element, it should be understood that the other element does not exist in the middle.

Terms such as first, second, etc. may be used to describe various elements, but the elements should not be limited by the terms. The above terms are used only for the purpose of distinguishing one component from another.

The terms used herein are used only to describe specific embodiments, and are not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly dictates otherwise. In this specification, terms such as "comprises" or "have" are intended to designate that the features, numbers, steps, operations, components, parts, or combinations thereof described in the specification exist, but one or more other features It should be understood that this does not preclude the existence or addition of numbers, steps, operations, components, parts, or combinations thereof.

In addition, the components of the embodiment described with reference to each drawing are not limitedly applied only to the embodiment, and may be implemented to be included in other embodiments within the scope of maintaining the technical spirit of the present invention, and also Even if the description is omitted, it is natural that a plurality of embodiments may be re-implemented as a single integrated embodiment.

In addition, in the description with reference to the accompanying drawings, the same components regardless of the reference numerals are given the same or related reference numerals, and the overlapping description thereof will be omitted. In describing the present invention, if it is determined that a detailed description of a related known technology may unnecessarily obscure the gist of the present invention, the detailed description thereof will be omitted.

In addition, terms such as “…unit”, “…unit”, “…module”, “…group”, etc. described in the specification mean a unit that processes at least one function or operation, which is hardware or software or hardware and software. It can be implemented by combining

1 is a block diagram of a big data-based online shopping mall AI keyword recommendation system according to an embodiment of the present invention, FIG. 2 is a block diagram of a big data collection module, and FIG. 3 is a distributed storage of data in the big data collection module is a block diagram of a part for, FIG. 4 is a diagram showing a sharding structure for data distributed storage, FIG. 5 is a diagram showing a data distributed storage structure, and FIG. 6 is a big data based online according to an embodiment of the present invention. It is a flowchart of a shopping mall AI keyword recommendation method, FIG. 7 is a table showing a transaction database and frequent item sets for explaining a keyword recommendation process, and FIGS. 8 to 10 are each seller product page, top-ranked search product information, and image recognition A drawing showing the result screen.

The existing e-commerce solution field is focused on implementing a system that integrates management of products sold, order management, and customer management. Since third-party e-commerce solution providers do not have big data on customer activities (search history, purchase patterns, etc.) only it was possible

The market demand is changing from a solution field that efficiently stores and manages big data collected from the existing e-commerce platform to a sales marketing analysis solution field that can directly increase sales by activating sales. Data marketing analysis technology that focuses on revitalizing sales based on big data such as self-collected customer activity search and purchase patterns, centering on major e-commerce platform providers, is growing.

In addition, AI open market, a growing field of e-commerce, provides a big data-based sales marketing analysis function that allows platform providers to provide their own excellent sales management system and increase seller sales.

However, since the AI open market is mainly managed by sellers with high sales and at the same time focusing on the implementation of marketing functions on the consumer side, small business sellers must market themselves in a sales method optimized for the sales channel platform technology.

The big data-based online shopping mall AI keyword recommendation system and method according to an embodiment of the present invention is not a platform-based recommendation system, but a seller product-based recommendation system, and a seller (seller) who wants to sell any product in an open market It is characterized by optimizing seller products as service users and recommending them to the platform. The open market is the main market sector, and it can induce an increase in inflow to new customers through product exposure.

Referring to FIG. 1 , the big data-based online shopping mall AI keyword recommendation system 100 according to the present embodiment includes a big data collection module 110 , a big data analysis module 120 , and a customer channel module 130 .

The big data collection module 110 collects data on the customer movement path, such as customer purchasing activities and search histories shopping on the open market platform. The big data collection module 110 may periodically collect a purchase pattern history, related search product information (photo/video, product information, etc.), related search terms, and the like among data based on a customer's shopping activity.

The big data collection module 110 may include a main execution unit (Execute Main), a foundation unit (Foundation), a processing unit (Process), and an element unit (Element) (see FIG. 2 ).

The Execute Main part is an introduction to the beginning of ESSD (Ecommerce Shopping Scheduler Daemon), and is responsible for setting the environment for execution of the upper element, configuring the API required for the ESSD element, and driving the process.

Execute Main part includes XML parser, DB pool manager, log controller, IPC manager, process controller, load environment module, etc. Its main functions include initialization of environment variables, process operation, community server operation, and schedule operation. It also includes a function to record and save process results. Foundation is initialized by Execute Main, and includes the interface (IPC Community, Socket Server) part between the processes in the upper area, the Utility API in the upper area, and the APIs necessary during the operation of the upper program such as the Log API to store detailed records such as operation status. do. The Foundation part includes a log module, an IPC community module, a utility module, and a web drive loader.

Process is driven by Execute Main, and the number and type of operation (one-time or independent program of Daemon type) are also determined by Execute Main. The Process part includes a One life cycle process module and a Daemon process module. Process is responsible for driving the upper element and delivering the result value to Execute Main. The selection of the type of element operation and the type of operation is received from Execute Main, and communication is communicated through the IPC Community of the Foundation area.

Element includes Web Crawler engine, Web Data Parser, specific Web data reading and Web Input automation functions. Element is driven by receiving information about Web URL, function type, and parser by the process, and the data corresponding to the result is delivered to the process area through the IPC community in the foundation area.

The Element Web Crawler engine is composed of Selenium + JSoup, and each library is included. Using Selenium Library, you can have a function to collect information on a specific web screen of a shopping mall and an action function within the web screen. The action function can be macroized by using a web browser driver to perform functions such as input label, button click, and scroll movement of the web screen. The crawling function may be a function of extracting necessary data by parsing HTML tags from a web screen.

Since the data that the big data collection module 110 stores in real time or periodically is large-capacity, a large-capacity distributed data storage technology may be applied.

Referring to FIG. 3 , the big data collection module 110 may include an ESSD server 111 , a database server 112 , and a web server 113 . Each server can be configured as a cloud system.

The ESSD server 111 is a batch server, and is a server on which the ESSD program operates. The product information and product ranking information of the shopping mall operator are obtained from the shopping mall site and stored in the database server 112 . In addition, when the web server 113 requests the ESSD function, corresponding data may be delivered.

The database server 112 may include a database (eg, MongoDB) for storing big data, and a database management unit (eg, MySQL) for storing shopping mall operator information, product settings for each operator, and the like.

The web server 113 is a web service server to be displayed to a shopping mall operator. A My Product screen that displays product price comparison, product ranking, number of visitors, keyword suitability, order status, etc. and an administrator screen for managing shopping mall sellers can be configured as services.

Referring to FIG. 4 , the big data collection engine 110 may set up a database to store large-capacity data and further configure the database through sharding settings.

For big data storage, the database server 112 may be configured through the sharding function of MongoDB. Referring to FIG. 5 , a data distributed storage structure is shown. The Config server is set in three clusters, the mongos server and the main mongod server are configured in one server, and additional servers can be configured as new servers. And you can pre-set the settings for additional configuration.

Referring back to FIG. 1 , the big data analysis module 120 analyzes the big data collected by the big data collection module 110 .

The big data analysis module 120 may include a data mining unit 121 , an image recognition unit 122 , and a learning analysis unit 123 .

The data mining unit 121 analyzes big data through data mining. One or more of DCA, SLM, and ARD analysis algorithms may be applied to big data analysis.

DCA (Data Cluster Analysis, Data Cluster Analysis) causes data points to form a cluster when there is an unknown characteristic in the data collected by the big data collection module 110, and groups them based on the similarity of these data characteristics (clustering). ) is a technique to create grouping data based on distribution within the group again.

SLM (Single Linkage Method) is a method of grouping two data into a large cluster with high similarity according to the shortest distance among the cluster distances between the two data.

ARD (Association Rule Discovery) is a method of discovering relevant keywords in relationship data between shopping customers and commerce products, and extracting recommended keywords that can optimize exposure of seller's products.

The image recognition unit 122 classifies an image related to a product and reads an object. And the keywords of the objects in the image are derived. The image recognition unit 122 may be provided outside, for example, may utilize an image recognition API (image recognition API) according to computer vision provided by Google Cloud.

The learning analysis unit 123 analyzes the keywords derived from the data mining unit 121 and the image recognition unit 122 for each seller product registered by the user of the present system through machine learning. In relation to the seller products registered by the user, the search ranking by keyword of customers' shopping search trends is tracked and analyzed on the e-commerce platform (AI open market).

The customer channel module 130 recommends an exposure optimization keyword for each seller product. If the exposure optimization keyword is reflected in the product name or product tag, the search ranking may rise and the number of visitors may increase after a certain period of time (1 to 3 days) has elapsed depending on the degree of product optimization.

To this end, the customer channel module 130 may include a keyword recommendation unit 131 .

The big data collection module 110 may first collect search-related keywords for each category of open market products, popular trend keywords for each product, and each trend keyword. In addition, keywords extracted from product images registered by users, product description keywords, and product tag keywords can be used for secondary collection. You can track the exposure frequency and ranking (search ranking) of seller products based on the keywords extracted by matching the first collected keywords with the second collected keywords.

By quantifying the exposure frequency and search ranking based on the extracted keyword of the seller product, the seller can derive a keyword that is not set in the seller product and provide it to the seller as an optimized keyword for each product. In this case, it is also possible to provide a trend in the ranking (search ranking for each product keyword by the seller) of products being exposed for each search keyword.

If the seller adds or modifies the recommended exposure optimization keyword to the product name, detailed description, tag, etc., the number of searched exposure keywords may increase and the ranking may rise. In this embodiment, the customer channel module 130 may automatically add or modify one or more of the product name, detailed description, and tag for the recommended exposure optimization keyword.

In addition, the customer channel module 130 may include a trend prediction unit 132 . The trend prediction unit 132 may predict popularity for each product category and recommend a user preferred product group for each category (popularity prediction keyword trend for each product category) in the open market. In addition, when registering a new product, product groups (keywords) that are optimized for the open market can be recommended together in a short period of time. When registering a product with the recommended keyword feature, the exposure rate can be significantly increased.

Also, the trend prediction unit 132 may monitor/analyze products of a shopping mall designated by the user as a competitor to provide a search trend for each product.

Also, the customer channel module 130 may include a product management unit 133 . The product management unit 133 may notify the service user of a report on optimization issues (ranking change, inflow change, etc.) for the seller product registered by the user in a predetermined manner (SMS sending, mobile web page linkage, etc.).

The product management unit 133 may provide inventory management and order management for seller products. Through the order status dashboard, it is possible to provide waiting status for each customer and order statistics for each product.

In addition, the product management unit 133 may manage the lowest price for the price comparison product. It is possible to support price comparison ranking, release counterparty, and lowest price automation within products tied to price comparison.

Referring to FIG. 3 , a method for recommending AI keywords for online shopping malls based on big data will be described. The method may be performed by the learning analysis unit 123 and the keyword recommendation unit 131 .

In order to use the big data-based online shopping mall AI keyword recommendation solution, keywords (tag keywords) matched by service users registered as customers for seller products are collected (step S200). Also, a combination keyword combining a product name and a category may be included as a reference meta tag.

For each reference meta tag, an extended keyword may be derived by using a related search word, an auto-completed search word, and a shopping related search word provided by a portal site and/or a shopping mall site.

Among the product name/tag keyword for each product category and the search product name/tag keyword for each keyword, it is possible to collect and include a product tag that is frequently used, ie, as a related keyword.

Also, a keyword of a current product that is searched and exposed in a tag keyword of a related product may be included as a keyword related to the related product.

Posts (blogs) registered by SNS users who have more than a certain number of subscribers (followers) on social network services (SNS) such as power blogger and Instagram/Facebook that are searched when a search is performed on the portal site using these various keywords , photos, texts, etc.) can be derived as influencer keywords.

In addition, words extracted from videos registered by video uploaders (one-person broadcasters, YouTubers) with more than a certain number of subscriptions searched on video sharing sites (eg, YouTube, etc.) using these various keywords are used as influencer keywords. may be derived. In the case of an image, words included in text displayed on the image may be extracted through image analysis, or words included in an explanatory voice of a video uploader included in the image may be extracted through voice analysis.

In this case, the more recent the posting or video registration date of an influencer (power blogger, SNS user with more than a certain number of subscribers, video uploader with more than a certain number of subscriptions, etc.) By applying a relatively high weight to the words derived from the video, it is possible to increase the possibility of extracting the influencer keyword reflecting the trend.

These various keywords (extension keywords, related keywords, product related keywords, influencer keywords) may be filtered by removing keywords not related to the category of the seller product or removing duplicates.

The results of searching in the open market are analyzed using the keyword population including the collected keywords (step S210).

As a result of the analysis, a recommended keyword indicating a high ranking of the seller product is selected and recommended to the service user (step S220).

When selecting keywords in step S220, keywords indicating a high ranking of similar products sold by other sellers, not products sold by the seller, may be selected and recommended to the seller. The criteria for calculating recommended keywords are as follows. Hereinafter, each numerical value is exemplary and may be changed as necessary.

A maximum of 30% can be given depending on whether or not the search is exposed on the basis of an additional 3% for each 10 keywords. 3% is added for each keyword whose ranking is exposed among 10 tag keywords. If all 10 are exposures, it can be 30%.

A maximum of 30% can be given depending on the ranking based on an additional 3% for each 10 keywords. If the ranking among the exposed keywords is within the designated ranking range (eg, 1, 2P), it is considered to be applicable. You can add 3% for each keyword. And depending on the number of impressions and clicks, it can be added within the range of 20%. 10% for 100-200 clicks (inflows) per product, 15% for more than 201 clicks (inflows). 5% for more than 5000 impressions per product, 0% for less than 5,000 impressions per product.

And depending on the number of orders, it can be added within the range of 20%. If the order quantity per product is 10-20, 10% is given, less than 10 is 0%, and if more than 21, 20% is given.

If all of the above conditions are satisfied, it becomes 100%, and the higher the %, the higher the keyword optimization rate for each product. The numerical standard of each item is daily data and may be data from the previous day.

An association analysis method may be applied during data mining in the data mining unit 121 . Association analysis is to discover association rules from a set of data including transactions or events and to calculate the correlation between two or more items.

For this purpose, association rule analysis is performed, and how frequently two sets of different items occur (relevance degree) can be derived.

There is transaction data for searching for association rules, and you can set the minimum support.

As shown in (a) of FIG. 7 , it is assumed that transaction data is given and the minimum support is set to 3.

Assuming the purchase data of the shopping mall, {1,2,3,4} item 1, item 2, item 3, and item 4 were purchased together, and this can be viewed as one item set. .

First, the first set of frequent items is found in the above transaction. Referring to (b) of FIG. 7 , each item has been purchased 3 times, 6 times, 4 times, and 5 times, and since they all have a minimum support rating of 3 or more, they may be included in the first frequent item set.

Next, a second frequent item set is obtained based on the first frequent item set. In this case, self-join and pruning may be performed.

The self-combination process is a process of generating all candidates of length 2 that can be generated based on the items {1}, {2}, {3}, and {4}. {1,2}, {1,3}, {1,4}, {2,3}, {2,4}, {3,4} may be generated.

Pruning is performed on the candidate generated in this way, and the pruning target may be a target to be pruned if there is an item that is not in the first frequent item set among the items of the item set among the candidates.

Items of each item set of {1,2}, {1,3}, {1,4}, {2,3}, {2,4}, and {3,4} are all included in the first frequent item set Therefore, there are no pruned candidates here.

Candidates that have undergone self-binding and pruning are counted as targets (see FIG. 7(c)). A second frequent item set is obtained from item sets in which the counted value is greater than the minimum support level of 3. That is, the {1.2}, {2,3}, {2,4}, and {3,4} item sets become the second frequent item set. {1,3}, {1,4} are no longer used in the next stage of frequent item set analysis. This is because if one itemset is infrequent, all sets including the item are also infrequent.

Next, a third frequent item set is obtained based on the second frequent item set. This is similar to when finding the second set of frequent items. Self-combining and pruning are performed on the second frequent item set.

Self bonding: {1,2}, {2,3}, {2,4}, {3,4}

--> {1,2,3}, {1,2,4}, {1,3,4}, {2,3,4}

Pruning: The subset of {1,2,3}, {1,2,4}, {1,3,4}, {2,3,4} is {1,2}, {1,3}, {2,3}, but {1,2,3} is pruned because {1,3} does not exist in the second frequent item set. {1,2,4}, {1,3,4} are also pruned in principle.

The transaction database is scanned for candidates {2,3,4} with a length of 3 and counted (refer to (d) of FIG. 7).

Although it is necessary to obtain the third frequent item set with the item sets whose counted value is greater than the minimum support 3, it may be terminated because there are no more item sets that satisfy the minimum support.

A first set of frequent items is obtained while scanning the transaction database.

A (k+1)th frequent item set is obtained for the kth frequent item set. In this process, candidates are found through self-combination and pruning, and only candidates that satisfy the minimum support condition are derived while scanning the transaction database.

The above process is repeated until the (k+1)th frequent item set is no longer created.

If the degree of frequent grouping between items is derived through this, the degree of association between each item may be calculated. Relevance can be confirmed through support, confidence, and lift.

Support = frq(X, Y) / N

Confidence = frq(X, Y) / frq(X)

Lift = Support / (Supp(X) x Supp(Y))

where frq(X, Y) is the number of times item X and item Y came out of the entire database together, frq(X) is the number of times item X came out of the entire database, and frq(Y) is the number of times item Y came out of the entire database It is the number of times, and Supp(X) and Supp(Y) are the probabilities of item X and item Y, respectively.

In this case, the degree of association (Rule: {1} -> {2}) from item 1 to item 2 in the transaction database shown in FIG. 7(a) is obtained as follows.

Support: Probability of 1 and 2 coming together in the entire database :: 3/7

Reliability: Probability of getting 2 when 1 is rolled: 3/3

Improvement: Assuming independence between items 1 and 2, P(A, B) / P(A)*P(B) :: 7/6

The data mining unit 121 may collect keywords having high relevance through the calculated relevance.

Hereinafter, the keyword recommendation algorithm will be described with drawings and examples for the convenience of understanding and description of the invention.

The following categories, product names, and tags may be set for the current seller product (see FIG. 8 ).

Category : Fashion Clothing > Women's Clothing > Coat

Product Name : [SALE] Winter Slim Women's Wool Cord (Same Day Delivery)

Tags: #wool code #winter women's coat #women's cork #women's coat #women's wool coat #women's slim coat #winter coat #small woman's coat #woman's winter coat

In this case, in step 1, it is checked whether there are the same characters as the category name in the product name. Only the product name is used as the basic tag (keyword) only if there is, and if not, the product name + category name is used as the basic tag.

In step 2, search for related keywords from autocomplete, related search terms, and representative categories on the portal site (eg Naver).

In step 3, the top 5 tag information of the product list output from each product representative category is collected using the search word and product name. In addition, the top 5 tag information of the product list output from each product detailed category is collected by using the search word and product name. In this case, the top 5 products with item tag information may be selected (see FIG. 9 ).

In step 4, image analysis information is collected through the image recognition unit 122 or through the API of an external image recognition device (eg, Google) using the registered product image (see FIG. 10 ).

In step 5, the same tag (keyword) is removed.

In step 6, tags are extracted from the conditional search results for each keyword. In this case, you can use the API provided by the open market. For example, in the case of the Naver Shopping API, https://developers.naver.com/docs/search/shopping/ is the same.

By using the final tag (keyword) finally derived in step 7, the priority of the keyword for the corresponding product in the open market is determined.

In this embodiment, the big data-based online shopping mall AI keyword recommendation method may also be implemented in the form of a recording medium including instructions executable by a computer, such as an application or program module executed by a computer. Computer-readable media can be any available media that can be accessed by a computer and includes both volatile and nonvolatile media, removable and non-removable media. In addition, computer-readable media may include computer storage media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data.

The above-described method may be executed by an application basically installed in the terminal (which may include a program included in a platform or operating system basically installed in the terminal), and the user may use the application store server, application, or related service It may be executed by an application (ie, a program) installed directly on the master terminal through an application providing server such as a web server. In this sense, the above-described method may be implemented as an application (ie, a program) installed basically in a terminal or directly installed by a user and recorded in a computer-readable recording medium such as a terminal.

Although the above has been described with reference to the embodiments of the present invention, those of ordinary skill in the art can variously modify the present invention within the scope without departing from the spirit and scope of the present invention described in the claims below. and may be changed.

100: Big data-based online shopping mall AI keyword recommendation system
110: big data collection module 120: big data analysis module
121: data mining unit 122: image recognition unit
123: learning analysis unit 130: customer channel module
131: keyword recommendation unit 132: trend prediction unit
133: product management department

Claims (4)

As a system that recommends keywords according to artificial intelligence (AI) in online shopping malls using big data,
a big data collection module that collects big data about seller products sold on the open market platform;
a big data analysis module that analyzes the big data and performs keyword analysis on the seller product; and
a customer channel module for recommending an exposure optimization keyword for the seller product based on the keyword analysis result,
The big data analysis module collects tag keywords matched for the seller product as a reference meta tag,
AI keyword recommendation system for online shopping based on big data, characterized in that the keyword analysis is performed by analyzing the posts or images of the found influencers using the reference meta tag and deriving them as influencer keywords.
According to claim 1,
Characterized in that at least one of power bloggers searched on portal sites, SNS users who have more than a certain number of subscribers or followers on social network services, and video uploaders with more than a certain number of subscriptions searched on video sharing sites are found as influencers. Big data-based online shopping mall AI keyword recommendation system.
3. The method of claim 2,
The more recent the registration date of the post or video of the influencer, the higher the number of views of the post or video, the higher the weight is applied to the word derived from the post or video to be derived as the influencer keyword Big data-based online shopping mall AI keyword recommendation system, characterized in that.
According to claim 1,
Extracting a word included in the text displayed in the video through image analysis from the video or deriving a word included in the description voice of the video uploader included in the video through voice analysis as the influencer keyword Big data-based online shopping mall AI keyword recommendation system.

Images