KR101026544B1 - Method and Apparatus for ranking analysis based on artificial intelligence, and Recording medium thereof - Google Patents

Abstract

A ranking analysis method based on artificial intelligence, a recording medium recording the same, and an apparatus thereof are disclosed. The present invention relates to a ranking analysis method based on artificial intelligence. In a method for analyzing a property information for each object of an information providing page by a ranking analysis device connected to a communication network by wire or wireless, a plurality of attributes corresponding to each of a plurality of objects. Establishing a relationship between the plurality of attributes by a ranking analysis device that defines the attributes; Measuring a value of each of the plurality of attributes and calculating a value for a relationship between the plurality of attributes for a user accessing the informational page; Weighting the value calculated for the relationship by performing a time analysis based on at least one of a user's stay time or a current time for each object of the information providing page; And extracting a clustering class by applying the weighted value to a clustering algorithm. According to the present invention, it is possible to provide an inference engine that can extract in real time an object whose property is matched to a tendency that changes with a user's time in a web environment. Analyze individual purchasing tendencies or suggest products that are tailored to your personal preferences, target your ads to market conditions in real time, and organize online-based advertising to suit your personal preferences or behaviors. The cost of preparing information for product production and marketing in the fields of digital content and advertising can be reduced.

Description

Ranking analysis method based on artificial intelligence, recording medium recording the same, the device {Method and Apparatus for ranking analysis based on artificial intelligence, and Recording medium}

The present invention relates to real-time statistical data analysis, in particular, ranking analysis method based on artificial intelligence that can extract in real time the object of the attribute that matches the tendency to change over time in a web environment, a recording medium, recorded therein, To the device.

Since the emergence of the Internet in recent years, the Internet users have rapidly increased and become very popular due to the development of information and communication technology, the spread of computers, and the development of Internet-based technology. In this situation, many companies open websites on the Internet to promote and provide information to netizens around the world, and e-commerce, which directly trades goods and services, is also becoming very common. Companies are working hard to devise new ways to provide information, service and revenue in a more convenient way for their users.

Electronic commerce is a transaction in which all or part of the transactions of goods and services are processed by electronic means such as the exchange of electronic documents and information. E-commerce using the Internet requires a distribution process that is required for the existing commerce to distribute products to consumers and generate profits, a sales network to identify customer demand, space for displaying and selling products, and management. Unlike the local, temporal and spatial limitations, which require a large number of human resources, commerce is carried out in the virtual space of the Internet, which can dramatically reduce the distribution process and distribution costs by transcending local, temporal and spatial limitations. Has an effect.

Generally, a marketing system using a database or database about statistics is prepared by a manager who fills out a questionnaire, aggregates it, and enters the content to build the content.The consumer inputs the information directly by the manager to search and inquire and respond to it manually. I wear it. In addition, the database that performs statistics and analysis also stores and outputs the structure and outputs the results of statistics only, and does not respond to artificial intelligence such as presenting the appropriate product to the individual or analyzing consumer demand and trend analysis. In addition, the fashion field has a lot of different emotions and numerous cases in which personal tastes, inclinations, and goods are used, and in order to respond to consumers by passively responding to them, managers do not have much information and knowledge on their own. Not only is it impossible, but it also requires a lot of manpower. In some cases, it is simply a database that is prepared for the number of cases by emotion item, but this is very difficult to apply to the prediction of outcomes in the fashion field, where the consumer's tendency is diverse and there is a great possibility of error.

In particular, the current statistical system collects data through standardized responses such as questionnaire surveys outside of the system, and questionnaire about human emotion that is not expressed in a specific language is impossible at all. Accordingly, in the fashion or digital content field where purchases are made based on individual emotions, it is very difficult to analyze individual purchase tendencies or propose products suitable for individual tendencies. In addition, it is time-consuming and costly to conduct continuous surveys to respond to the rapidly changing market of fashion and digital contents every season, and the standardized surveys are difficult to grasp the emotions of consumers. There is nothing else. Therefore, at present, the provision of information for the production and sale of products in the field of fashion or digital contents is completely dependent on the method by the prediction of professional personnel.

Meanwhile, advertising through interactive media has become popular in recent years. As the number of people using the Internet has exploded, advertisers have become aware of the media and services offered over the Internet as a potential and powerful way for advertising. Advertisers have developed several strategies in an attempt to maximize the value of advertising. The first strategy is for advertisers to use websites with many users as a means of providing interactive media or services. In this case, the advertiser may place ads on the home page of the New York Times website, or the USA Today website, for example. In the second strategy, advertisers increase their likelihood of positive response by the user by targeting their ads to users in a narrower niche. For example, an agency promoting tourism in the Costa Rica rainforest may place advertisements on the Ecotourism-Travel subdirectory of the Yahoo Web site. Advertisers typically go through this targeting process manually. These methods are not adaptable to changing market conditions in real time and require a large number of skilled workers. In addition, even though advertisers can reach many users in website-based advertising, they are often dissatisfied with the return on their advertising investment. This is because traditional online-based advertising is far from analyzing an individual's disposition or behavior.

The first technical problem to be achieved by the present invention is to provide a ranking analysis method based on artificial intelligence that can extract in real time the object of the attribute matching the tendency to change over time of the user.

The second technical problem to be achieved by the present invention is to provide a ranking analysis apparatus based on artificial intelligence to which the ranking analysis method based on the artificial intelligence is applied.

In order to achieve the first technical problem, the present invention provides a method for analyzing a property information for each object of an information providing page by a ranking analysis device connected to a communication network by wire or wireless, wherein a plurality of properties corresponding to each of a plurality of objects are selected. Establishing a ranking analysis device defining a relationship between the plurality of attributes; Measuring a value of each of the plurality of attributes and calculating a value for a relationship between the plurality of attributes for a user accessing the informational page; Weighting the value calculated for the relationship by performing a time analysis based on at least one of a user's stay time or a current time for each object of the information providing page; And extracting a clustering class (simply referred to as a 'class') by applying the weighted value to a clustering algorithm. Here, the plurality of objects may include at least one of digital content, product, or advertisement content.

In the ranking analysis method based on artificial intelligence according to an embodiment of the present invention, before the clustering class is extracted, the behavior pattern of the user is extracted by performing behavior analysis based on the response information of the user in the information providing page. It may further comprise a step. In this case, extracting the clustering class may include data mining the weighted value based on the behavior pattern of the user.

Here, the response information may include login information of the user, information about an object read by the user, information about an object purchased by the user, information about the number of purchases of the user, time information for each object, or information input by the user. It may include at least one of the post content.

In addition, the behavior pattern may include at least one of the user's interest, purchasing power, purchase propensity, main activity time, purchase time, preferred brand, or preferred weather.

In the ranking analysis method based on artificial intelligence according to another embodiment of the present invention, in the step of extracting the behavior pattern of the user, the ranking analysis apparatus may extract the behavior pattern of the user based on predefined behavior analysis metadata. have.

In a ranking analysis method based on artificial intelligence according to another embodiment of the present invention, in the step of establishing a relationship between the plurality of properties, the relationship may be set based on relationship history data in which a relationship previously set between the properties is recorded. Can be.

According to another embodiment of the present invention, a ranking analysis method based on artificial intelligence is calculated based on time history data including at least one of accumulated time or activity time accumulated for the user in assigning the weight. The method may include determining a weight to be assigned to the value.

In a ranking analysis method based on artificial intelligence according to another embodiment of the present invention, prior to the step of establishing a relationship between the plurality of attributes, a ranking analysis apparatus that defines a plurality of attributes corresponding to each of the plurality of objects is queried. The method may further include receiving an input. In this case, in the extracting of the clustering class, the clustering class corresponding to the query may be extracted.

In order to achieve the second technical problem, the present invention defines a plurality of attributes corresponding to each of a plurality of objects, and is connected to a communication network by wire or wireless, in the apparatus for analyzing the attribute information for each object of the information providing page An attribute matching unit for setting a relationship between the plurality of attributes; A value measuring unit measuring a value of each of the plurality of attributes and calculating a value for a relationship between the plurality of attributes for a user accessing the information providing page; A time analyzer for weighting the value calculated for the relationship by performing a time analysis based on at least one of a time spent by the user for each object of the information providing page or a current time; And a clustering unit extracting a clustering class by applying the weighted value to a clustering algorithm.

The ranking analysis apparatus based on artificial intelligence according to an embodiment of the present invention may further include a behavior analysis unit configured to extract behavior patterns of the user by performing behavioral analysis based on the response information of the user in the information providing page. Can be. In this case, the clustering unit may data mining the weighted value based on the behavior pattern of the user.

Here, the reaction information may include login information of the user, information about an object read by the user, information about an object purchased by the user, information about the number of purchases of the user, time information for each object, or input by the user. It may include at least one of the contents of a post.

According to another embodiment of the present invention, a ranking analysis apparatus based on artificial intelligence may be assigned to the calculated value based on time history data obtained by recording at least one of accumulated time or activity time accumulated for the user in the time analyzer. The weight can be determined.

An artificial intelligence-based ranking analysis device according to another embodiment of the present invention stores a relationship set by the attribute matching unit as relationship history data and includes at least one of accumulated time or activity time accumulated for the user. The apparatus may further include a data storage module that stores history data and stores the extracted clustering class.

The ranking analysis apparatus based on artificial intelligence according to another embodiment of the present invention may further include a query analyzer configured to receive a query and perform morphological analysis. In this case, the clustering unit may extract a clustering class corresponding to the morphologically analyzed query.

According to the present invention, by analyzing the relationship between the user's propensity and various objects in the web environment and assigning weights by time analysis to the analyzed result, the object of the attribute matching the user's time-changing tendency is extracted in real time. Inference engine can be provided, and when using such inference engine, it is possible to analyze personal purchase tendency in the field of fashion or digital content, suggest a product suitable for individual tendency, and target the advertisement in real time market. It is possible to perform according to the situation and to configure the online-based advertising to suit the individual's inclination or behavior, and to reduce the cost of preparing information for product production and marketing in the fields of fashion, digital content, and advertising.

Hereinafter, with reference to the drawings will be described a preferred embodiment of the present invention. However, embodiments of the present invention illustrated below may be modified in various other forms, and the scope of the present invention is not limited to the embodiments described below.

1 is a block diagram of a ranking analysis apparatus based on artificial intelligence according to an embodiment of the present invention.

External statistical data collection unit 1 to 3 (111-113) collects data related to the user's login information, the object selected or purchased by the user. The data collection module does not need to be embedded in the ranking analysis device, but may be configured as a separate relay device on a network connected to the Internet. The external statistical data collection units 1 to 3 (111 to 113) may be classified according to the type of the object. For example, each of the collection units may be optimized for products, digital contents, and advertisement contents.

The query analyzing unit 120 may receive a query command from the external statistical data collecting units 1 to 3 (111-113). The query analyzer 120 serves as an interface for connecting to other objects on the Internet. However, according to the needs of those skilled in the art, the query analyzer 120 may be omitted, and the behavior analyzer 130 may be used as an interface between devices, or the query analysis function may be merged into the behavior analyzer 130. The query analyzing unit 120 analyzes the input query (a query input from an external search engine, a query command received from the external statistical data collecting unit 111-113, etc.) and classifies the query contents. Through query analysis, an object (time, relationship, search, etc.) of the ranking analysis device to be searched is determined.

Here, the query command includes a ranking analysis device based on artificial intelligence and a code promised in advance according to an embodiment of the present invention. However, when interworking with Naver or other search engines, it may include unscheduled code or general sentences. In this case, the query analyzing unit 120 converts it into a form of a query command. For example, in the ranking analysis device, "I'm looking for Top products among the most popular items." Is entered, a simplified query command such as "recent fashion", "product top search", etc. may be derived.

The behavior analysis unit 130 analyzes the data collected by the external statistical data collection units 1 to 3 (111-113) by matching the predefined behavior analysis metadata.

The object attribute matching unit 140 measures and analyzes various attributes by matching each object with each other.

The value measuring unit 150 measures not only the value of the attribute itself of the object but also the value of the relationship between each attribute.

The time analyzer 160 reflects the time analysis result in the result of the value measuring unit 150 to give a weight over time.

The clustering unit 170 performs clustering according to the values of the weighted attributes output from the time analyzing unit 160.

The output unit 180 transmits the analysis result of the clustering unit 170 to external statistical data collection units 1 to 3 (111-113) and other objects on the Internet.

The data storage module 190 stores and manages history or metadata. The data storage module 190 may analyze the behavior of the behavior analysis unit 130, the clustering result (or clustering class) of the clustering unit 170 (eg, product analysis content corresponding to a query, content or advertisement analysis corresponding to a query, etc.). Contents).

External statistical data collection unit 1 to 3 (111-113) can know what kind of objects purchased by a particular user, but in the case of inquiring historical data is connected to the data storage module 190 of the ranking analysis device to be. The data storage module 190 receives the data from the external statistical data collection units 1 to 3 (111-113) and stores the result of the analyzed "object vs relationship".

2 is a flowchart of a ranking analysis method based on artificial intelligence according to an embodiment of the present invention.

First, data is received from an external statistical data collection unit or a search engine (S210). This process may be omitted in some cases.

Next, the ranking analysis device receives a query (S220). In this process (S220), the above-described query analysis may be performed. The ranking analysis apparatus defines a plurality of attributes corresponding to each of the plurality of objects using metadata. Here, the plurality of objects may be at least one of digital content, product, or advertisement content.

Next, the user's behavior pattern is extracted by performing a behavioral analysis based on the input query, the user's response information, etc. on the information providing page accessed by the user (S230). Informational pages include general web pages, documents on the network, link pages linking them, sites with mobile access, and the like. Here, the reaction information may include at least one of a user's login information, information on an object read by the user, information on an object purchased by the user, information on the number of purchases of the user, time spent on each object, or content posted by the user. Can be. The information providing page to be analyzed may be a specific online shopping mall, a specific portal site, a specific content providing site, or the like. The behavior pattern extracted according to the behavior analysis may be information including the user's interest, purchasing power, purchase propensity, main activity time, purchase time, preferred brand, or preferred weather.

Thereafter, a relationship between a plurality of attributes is set (S240). In this process (S240), the relationship may be set based on relationship history data for a relationship previously set between the attributes.

Next, for each user accessing the informational page, the value of each of the plurality of attributes is measured. In addition, the value of the relationship between the plurality of attributes is calculated using this (S250).

When the value measurement is completed, the user performs a time analysis based on various time data related to the user, such as the time stayed for each object of the information providing page or the current time, and weights the calculated value (S260). More specifically, a weight to be assigned to the value calculated above may be determined based on time history data including at least one of accumulated time for the user or past and present main activity time.

Next, the clustered class corresponding to the query is extracted by applying the weighted value to the clustering algorithm (S270). According to an embodiment of the present invention, when extracting the clustering class, the data may be mined based on the behavior pattern of the user to extract the clustering class closer to the taste of the user. According to another embodiment of the present invention, an object (eg, shopping item, movie content, drama content, home appliance advertisement content, etc.) corresponding to the clustering class may be extracted.

Thereafter, the clustering class is output (S280), and various results analyzed above are stored as history data (S290). These processes (S280, S290) is not an essential process, it can be omitted according to the needs of those skilled in the art.

3 is a detailed flowchart of the query analysis process S220 of FIG. 2.

First, dictionary metadata, which is stored in advance, is loaded (S321). In the dictionary, the meaning is defined in advance for several queries.

Next, a word is extracted from the query based on the loaded dictionary metadata (S322). In other words, instead of interpreting all of the query content, only meaningful words are found and interpreted. For example, you can remove meaningless words like a, the, and this / that from a query.

Next, the morpheme analysis for each word separated from the query (S323). For example, analyzing the query "perform the stemmed word for each separated word" results in "stem-word stemming".

Finally, natural language analysis is performed (S324), which is a process of interpreting the meaning of the query content for each word after morphological analysis.

In the query analysis process, organic linkage with pre-stored dictionary metadata is required. In addition, this requires natural language processing and morphological analysis. For example, a verb may be used to determine the "to do" of the Hanging analysis apparatus. If the verb contains "search, result, find", etc., the ranking analysis apparatus according to an embodiment of the present invention may return a search result. Also, nouns in the query can determine the "object", category, and relationship. For example, if a noun includes "goods, clothes, clothing, top," it is possible to search for top, user, sales, and hits. In addition, time related data may be determined through adverbs and adjectives included in the query. For example, if the query "recent", "trend", "recent", etc. are included, the analysis can be limited to the data within one month.

4 is a detailed flowchart of the behavior analysis process S230 of FIG. 2.

First, behavior analysis metadata is loaded (S431). Behavior analysis metadata refers to the results of behavior analysis that is defined or already classified. Here, behavior represents behavior patterns, tastes, and preferences of users or objects.

5 illustrates an example of behavioral analysis metadata. Behavioral analysis metadata can be obtained using an Experimental Analysis of Behavior.

Next, a new behavior is analyzed by performing a behavior analysis algorithm (S432).

When the behavior is analyzed, clustering is performed based on the behavior analysis metadata (S433). Here, the user's behavior pattern may be extracted based on predefined behavioral analysis metadata.

As a result of the clustering, if a new cluster is found, that is, if clustering cannot be performed in the existing clustering class, a new class is added (S435) and the behavior analysis algorithm is executed again (S432).

If the addition of the class is not necessary, the current clustering result is reflected in the existing behavior analysis metadata (S436). This is a way to set the metadata according to the behavior analysis in advance, and learn the ranking analysis device itself through an iterative clustering process and machine learning.

FIG. 6 is a detailed flowchart of the object attribute matching process S240 of FIG. 2.

First, an object property is analyzed (S641). For example, you can check attributes such as price, popularity, color, etc. for a particular shopping item.

Next, the measured data for each property is analyzed (S642). Here, the data measurement may be performed in the data collection unit outside the ranking analysis device.

Next, the relationship that can be associated with each property is checked and analyzed (S643). For example, when the attribute of the clothing is color, price, quality, recognition, etc., the relationship by attribute may be analyzed between color and price and relationship between color and quality. Or, if the user's attributes are attributes such as cheerfulness, prudence on quality, and irrelevant to price, the relationships by attributes may be analyzed such as cheerfulness and color, prudence and recognition on quality. On the contrary, if the user's information is missing from the attribute-specific relationship, the relationship between color and price and the relationship between color and quality can be analyzed. If the user's attributes, cheerfulness, quality, and price are irrelevant, the relationship with cheerful, prudent and price-independent products can be analyzed. 7 shows an example of metadata about an object, an attribute, and a relationship. Various properties can occur for each object, and the value of the relationship can be measured when the relationship between these properties is established. For example, if we consider sales by time / vendor / region as an element of value measurement, the relationship can be defined as “timely / vendor / regional sales”. Accordingly, the value of the relationship may be calculated as, for example, "revenue of product + region of user + vendor ID + time date".

Next, by matching the historical data and attributes of the past and analyzing which attributes are most suitable to be associated (S644).

When the relationship between the attributes is completed, the analyzed data is stored (S645). The analyzed data may be stored in the data storage module 190.

The process of FIG. 6 matches the most similar attributes when the attributes defined in the ranking analysis apparatus and the attributes of the object from an external module are different, but the attributes and the relations from the existing history data. It is the process of deciding whether to import and analyze existing data.

8A is a detailed flowchart of the value measurement process S260 of FIG. 2.

First, the data is classified by object (for example, "User, Product, Contents, Vendor, Location, Advertisement, Time", etc.), and the relationship for outputting data corresponding to the previously entered query is arranged. And the data accumulated by the external statistical data collection unit is inquired (S851).

Next, the historical data stored in the ranking analysis device is inquired (S852).

Next, using the inquired data, the necessary factors (average, variance, standard deviation, etc.) for using the basic algorithm for measuring the value are calculated (S853).

Thereafter, a probability density distribution is calculated using the above calculated element (S854), and a cumulative distribution is calculated accordingly (S855).

Next, the cumulative distribution is quantized to obtain a value for each attribute and a value for each relation between the attributes (S856 and S857). 8B and 8C show examples of probability density functions and cumulative distributions. The probability density function can be found using the variance and the mean. The cumulative distribution table can also be obtained from the probability density function. The result of quantizing the cumulative distribution is the value of the property or the relationship between the properties. Here, the value for the relationship between the attributes may be a simple sum or weighted sum of the values for each attribute.

Next, data mining is performed so that the predicted data most similar to the behavior analysis result of the user can be extracted through the calculated value data.

 For example, when a query for "list of products most popular with user A" is input to the ranking analysis apparatus, the query contents are identified through query analysis, and the relationship between attributes is determined. If the attribute is a user's disposition, interest category, product hit count, sales history, and purchase history of the interest category, the value of each attribute is measured and the value of the scarcity of each relationship is calculated. In addition, the results of these calculations can be grouped again to allow for multiple complex value measurements. When the relationship value measurement is completed, data mining can be combined to output a list to answer the final query.

9 is a detailed flowchart of the time analysis process S260 of FIG. 2.

First, various relationship data according to the current time are determined (S961). For example, calculate environment variables such as the time a user stays on the page, whether the current time is weekday, weekend, day of the week, or shared / holiday.

The time clustering metadata is loaded (S962), and the time history data that matched the measured value and the time in the past is loaded (S963). The time history data can be used when the relationship or attribute is not newly added.

For example, if you want to measure value related to the number of views of a product, you can weight the number of views as shown in Table 1 in total 100 points.

Cumulative views over last week 50 points 1 week ago 30 points 1 week cumulative views two weeks ago 10 points 1 week cumulative views 3 weeks ago 5 points Cumulative views from 4 weeks to 2 months 5 points Other cumulative views 0 points

As shown in Table 1, the weight of the hit can be measured differently according to time.

When the hit product or the hit content is output only as the sum of the total views, the product or the content that has the maximum number of views during the specific season may always be ranked as the top one. However, by using the time history data, it is possible to adaptively track the tastes of individuals that change from time to time.

In other words, if a hit product or hit content is output only with the total number of hits in the last 1-2 weeks, the product or content that had the maximum number of views in a particular season is pushed out of the rank after 2 weeks regardless of the ranking. The time history data can be used by the supplier of the product or the content to more accurately determine the product or the content that is likely to be sold.

Next, the necessary elements (average, variance, standard deviation, etc.) for time analysis are calculated (S964).

Next, the value measurement data is clustered by time zone according to the content determined in the clustering metadata (S965). For example, the current user can get a lot of church-related web surfing on weekends and a search for menus at dinner around 5 o'clock.

Thereafter, the weight is calculated for each clustering class (S966). For example, it can be calculated to have a higher weight on common variables such as the current time zone and today's day.

Next, the weight is quantized (S967), and the quantized weight is reflected in the value measurement data (S968). Accordingly, the time factor can be reflected in the value measurement data.

The process of FIG. 9 reflects the time analysis result in the already measured value measurement data. Even if you don't have accurate information about whether the current user is a business, church, or home, you can estimate the current situation by analyzing the user's behavior, and recommend the most appropriate data for the current time by reflecting the estimated result in the value measurement data. can do.

FIG. 10 is a detailed flowchart of the clustering process S270 of FIG. 2.

First, data is classified by classifying objects according to attributes (S1071).

Next, the measured value for each relationship is analyzed (S1072). The value analyzed reflects the weight of the time analysis.

Next, a clustering algorithm (eg, K-Nearest Neigbor (KNN) algorithm, Naive Bayesian, etc.) is executed to summarize similar analysis results of the measured values for each relationship (S1073).

Thereafter, the result of the clustering algorithm is stored (S1074), and the final clustering class is extracted through the summarized result (S1075).

The invention can be implemented via software. Preferably, a ranking analysis method based on artificial intelligence according to an embodiment of the present invention may be provided by recording a program for executing in a computer on a computer-readable recording medium. When implemented in software, the constituent means of the present invention are code segments that perform the necessary work. The program or code segments may be stored on a processor readable medium or transmitted by a computer data signal coupled with a carrier on a transmission medium or network.

Computer-readable recording media include all kinds of recording devices that store data that can be read by a computer system. Examples of the computer readable recording medium include ROM, RAM, CD-ROM, DVD 占 ROM, DVD-RAM, magnetic tape, floppy disk, hard disk, optical data storage, and the like. The computer readable recording medium can also be distributed over network coupled computer devices so that the computer readable code is stored and executed in a distributed fashion.

Although the present invention has been described with reference to one embodiment shown in the drawings, this is merely exemplary, and it will be understood by those skilled in the art that various modifications and variations may be made therefrom. And, such modifications should be considered to be within the technical protection scope of the present invention. Therefore, the true technical protection scope of the present invention will be defined by the technical spirit of the appended claims.

The present invention relates to a ranking analysis method based on artificial intelligence that can extract in real time an object whose attribute is matched to a user's propensity to change in a web environment, a recording medium recording the same, an inference engine, and an online device. It can be applied to shopping sites, content routers, content repeaters, portal servers, and the like.

1 is a block diagram of a ranking analysis apparatus based on artificial intelligence according to an embodiment of the present invention.

2 is a flowchart of a ranking analysis method based on artificial intelligence according to an embodiment of the present invention.

3 is a detailed flowchart of the query analyzing process of FIG. 2.

4 is a detailed flowchart of the behavior analysis process of FIG. 2.

5 illustrates an example of behavioral analysis metadata.

6 is a detailed flowchart of an object attribute matching process of FIG. 2.

7 shows an example of metadata about an object, an attribute, and a relationship.

8A is a detailed flowchart of the value measurement process of FIG. 2.

8B and 8C show examples of probability density functions and cumulative distributions.

9 is a detailed flowchart of the time analysis process of FIG. 2.

10 is a detailed flowchart of the clustering process of FIG. 2.

Claims (16)

In the method of analyzing the attribute information for each object of the user interface or information providing page by a ranking analysis device connected to a communication network by wire or wireless, On at least one of the user's login information, information on the object read the review, information on the object purchased, information on the number of purchases of the user, time information for each object, or information posted by the user on the information providing page Extracting a behavior pattern including at least one of interest, purchasing power, purchase propensity, main activity time, purchase time, preferred brand, or preferred weather of the user by performing a behavior analysis that defines a behavior pattern based on the behavior; Setting a relationship between the plurality of attributes by a ranking analysis device defining a plurality of attributes corresponding to each of a plurality of objects; For the user, measuring a value of each of the plurality of attributes and calculating a value for a relationship between the plurality of attributes; Based on the current time, time analysis based on time history data including the user's stay time and past and present main activity time is performed for each object of the information providing page, and the weight of the relationship is given a different weight for each time zone. Making; And Clustering the plurality of objects based on the behavioral pattern of the user and the value for the relationship Including, ranking analysis method based on artificial intelligence. delete delete delete The method of claim 1, Extracting the behavior pattern of the user, The ranking analysis device is a step of extracting the behavior pattern of the user based on the predefined behavior analysis metadata, ranking analysis method based on artificial intelligence. The method of claim 1, The plurality of objects, Artificial intelligence-based ranking analysis method, characterized in that it comprises at least one of digital content, product or advertising content. The method of claim 1, Setting the relationship between the plurality of attributes, And setting the relationship based on relationship history data in which previously recorded relationships between the attributes are recorded. delete The method of claim 1, Before the setting of the relationship between the plurality of attributes, the ranking analysis apparatus further comprises the step of receiving a query, The clustering step, And extracting a clustering class corresponding to the query from the result of the clustering. A computer-readable recording medium having recorded thereon a program for executing the method of any one of claims 1, 5, 6, 7, or 9 in a computer system. An apparatus for defining a plurality of attributes corresponding to each of a plurality of objects, and connected to a communication network by wire or wirelessly to analyze attribute information for each object of a user interface or an information providing page, On at least one of the user's login information, information on the object read the review, information on the object purchased, information on the number of purchases of the user, time information for each object, or information posted by the user on the information providing page A behavior analysis unit configured to perform a behavior analysis that defines a behavior pattern based on the behavior pattern including at least one of the user's interest, purchasing power, purchase propensity, main activity time, purchase time, preferred brand, or preferred weather; An attribute matching unit for setting a relationship between the plurality of attributes; A value measurer for measuring a value of each of the plurality of attributes and calculating a value for a relationship between the plurality of attributes for the user; Based on the current time, time analysis based on time history data including the user's stay time and past and present main activity time is performed for each object of the information providing page, and the weight of the relationship is given a different weight for each time zone. A time analyzer; And A clustering unit for clustering the plurality of objects based on the behavior pattern of the user and the value of the relationship. Comprising, artificial intelligence based ranking analysis device. delete delete delete The method of claim 11, Storing the relationship set by the attribute matching unit as relationship history data, storing time history data including at least one of accumulated time or activity time accumulated for the user, and further comprising a data storage module. Ranking analysis device. The method of claim 11, It further includes a query analysis unit for receiving a query and morphological analysis, The clustering unit, And a clustering class corresponding to the query from which the morpheme is analyzed is extracted from the clustering result.

Images