KR20140031447A - Apparatus and method for classifying propensity of consumer - Google Patents

Abstract

Disclosed are an apparatus and method for classifying a consumer propensity. A consumer information collection unit receives, from a consumer who accesses an online site, consumer propensity information which is used to classify a consumer propensity according to predetermined consumer propensity classification, and collects a web shopping log which represents actions performed by the consumer to buy a product on the online site. A search log calculation unit calculates a search log which represents whether the consumer has searched for product related information which is previously set. An action log calculation unit calculates an action log based on the search log, wherein the action log represents whether the consumer has performed an action corresponding to predetermined consumer action types in response to the product related information. An activity index calculation unit calculates an activity index which represents a degree of required information search by a consumer to buy a product. A classification model generation unit creates a rule which is applied to classify a propensity of a consumer who accesses the online site based on the calculated action log, the activity index, and the consumer propensity information inputted by the consumer. A consumer propensity classification unit classifies a propensity of a consumer who accesses the online site based on the calculated action log, the activity log, and the generated rule. According to the present invention, service quality related to IT can be improved and a differentiated new kind of IT service can be provided by systematizing social scientific theories. [Reference numerals] (110) Consumer information collection unit; (120) Search log calculation unit; (130) Action log calculation unit; (140) Activity index calculation unit; (150) Classification model generation unit; (160) Consumer propensity classification unit

Description

Apparatus and method for classifying propensity of consumer}

The present invention relates to an apparatus and method for classifying consumer's propensity, and more particularly, to an apparatus and method for classifying consumer's propensity to reflect on various personalized services such as content recommendation, information search, and medical service. .

Recently, with the advent of various smart devices, as personal cloud services are provided, social interest in accumulation and utilization of integrated personal information is increasing. While most of the existing personalization services were limited to the field of content recommendation and information retrieval, in recent years, personalization services are actively performed in various fields including the medical field. In order to construct and provide such personalized services in various fields, it is necessary to consider factors influencing the formation of the user's disposition. Therefore, there is an increasing need for a personalization algorithm or a user modeling method that can minimize performance differences according to service providing fields.

The theory of consumer behavior in the social sciences suggests various theories to explain the causes of consumer behavior. In the theory of consumer behavior, the factors affecting consumer's decision making and consumption behavior are Internal Influences and External Influences. Internal factors include demographic information, lifestyle, motivation, personality, knowledge, attitudes, beliefs, feelings, etc. External factors include culture, region, and the Past Experience Reference Group.

Many of these consumption behavior factors are already being considered in data mining studies on personalized services. However, research on rather abstract concepts such as motivation, personality, attitude, belief, and feeling are among the early stages of internal factors.

In particular, the theory of consumer behavior suggests that people's internal factors, especially the propensity of people who belong to individuality, are important influence factors that influence consumer's decision making and consumption behavior. A representative theory of consumer behavior about human inclination is the Regulatory Focus Theory. The adjustable focus theory classifies a person's disposition into promotion focus and prevention focus. The focus of improvement is the propensity to have ideals, hopes, and aspirations as control mechanisms in pursuit of the goal, and the focus on prevention is the tendency to have obligations, responsibilities, and responsibilities as control mechanisms. These two propensities have a significant impact on consumer decision making in a variety of areas, such as information exploration, processing, and alternative evaluation methods. Therefore, the research to apply this to personalized service is attracting attention.

Prior document related to the present invention Enhancing Collaborative Filtering Systems with Personality Information (2011, Rong Hu) reflects the personality or personality of the individual in the recommendation system. This not only minimizes performance degradation due to cold start issues, but also improves the performance of existing collaborative filtering recommendation techniques.

However, there is a somewhat insufficient element to apply it directly to personalized services in that the correlation between personality and actual purchasing behavior of a person classified by the psychology personality discrimination tool has not been verified.

An object of the present invention is to provide an apparatus and method for automatically classifying consumer's consumption propensity to reflect various personalized services such as content recommendation, information search, and medical service.

Another technical problem to be solved by the present invention is to read a computer with a program for executing a method for automatically classifying consumer's consumption propensity to reflect in various personalized services such as content recommendation, information retrieval, and medical service. To provide a recording medium that can be.

In order to achieve the above technical problem, the consumer's disposition classification apparatus according to the present invention inputs consumer disposition information that can classify the consumer's disposition according to a preset disposition of the consumer from a consumer connected to an online site. A consumer information collection unit configured to receive and collect a web shopping log indicating an action performed by the consumer to purchase a product on the online site; A search log calculator configured to calculate a search log indicating whether the consumer searches for preset product related information; A behavior log calculator configured to calculate a behavior log indicating whether the consumer performs a behavior corresponding to a preset consumer behavior type in response to the product related information based on the search log; An activity index calculator for calculating an activity index indicating a degree to which the consumer searches for information necessary to purchase a product; A classification model generator for generating a rule for classifying a consumer's disposition to access the online site based on the calculated behavior log and activity index and consumer disposition information received from the consumer; And a consumer propensity classifier that classifies the consumer's propensity based on the behavior log and activity index calculated for the consumer accessing the online site and the generated rule.

In order to achieve the above technical problem, the consumer's propensity classification method according to the present invention inputs consumer propensity information for classifying the consumer's propensity according to a pre-set consumer's propensity classification from a consumer connected to an online site. A consumer information collection step of receiving and collecting a web shopping log indicating an action performed by the consumer to purchase a product on the online site; A search log calculating step of calculating a search log indicating whether the consumer searches for preset product related information; A behavior log calculation step of calculating, based on the search log, a behavior log indicating whether the consumer has performed an action corresponding to a preset consumer behavior type in response to the product related information; An activity index calculation step of calculating an activity index indicating a degree to which the consumer searches for information necessary for purchasing a product; A classification model generation step of generating a rule for classifying a consumer's propensity to access the online site based on the calculated behavior log and activity index and consumer propensity information received from the consumer; And a consumer disposition classification step of classifying the disposition of the consumer based on the behavior log and activity index calculated for the consumer accessing the online site and the generated rule.

According to the apparatus and method for classifying consumer tendencies according to the present invention, it is possible to improve the quality of IT-related services and provide a new kind of differentiated IT services by systemizing theories in the social sciences. In addition, it is possible to minimize the cold start problem, which is a chronic problem of the existing personalization algorithm, and to improve performance in terms of algorithm accuracy and speed.

1 is a block diagram showing the configuration of a preferred embodiment of a consumer's propensity classification apparatus according to the present invention;
2 is a view for explaining a process of receiving consumer tendency information that can classify consumer tendency according to a control focus theory;
3 is a table showing an example of a process of selecting important variables;
FIG. 4 is a graph showing the relationship between consumer preferences classified based on important variables selected through the process of FIG. 3 and consumer orientation information; FIG.
FIG. 5 is a diagram illustrating a decision tree generated based on an important variable selected through the process of FIG. 3;
6 is a flowchart illustrating a process of performing a preferred embodiment of the consumer's disposition classification method according to the present invention.

Hereinafter, exemplary embodiments of a consumer's propensity classification apparatus and method according to the present invention will be described in detail with reference to the accompanying drawings.

Figure 1 is a block diagram showing the configuration of a preferred embodiment of a consumer disposition classification apparatus according to the present invention.

Referring to FIG. 1, a consumer's disposition classification apparatus according to the present invention includes a consumer information collector 110, a search log calculator 120, a behavior log calculator 130, an activity index calculator 140, and a classification model. The generation unit 150 and the consumer orientation classification unit 160 is provided.

The apparatus for classifying consumer tendencies according to the present invention may classify consumer tendencies according to various consumer behavior theories in the social science field. Hereinafter, a process of classifying a consumer's disposition into a promotion focus or a prevention focus according to a regulation focus theory, which is one of consumer behavior theory, will be described. I will explain.

The consumer information collecting unit 110 receives consumer tendency information for classifying consumer tendencies according to a preset consumer tendency classification from a consumer accessing the online site. The classification of consumer tendency means to classify consumer tendency according to the theory of consumer behavior, which is a social science field. For example, when pre-set to classify the consumer's disposition according to the control focus theory, one of the representative consumer behavior theories, the consumer information collecting unit 110 may classify the consumer's disposition as an enhancement focus or a preventive focus. have.

Consumer propensity information input from the consumer refers to consumer-related information that can classify the consumer's disposition according to the consumer behavior theory defined as described above.

2 is a view for explaining a process of receiving consumer tendency information that can classify consumer tendency according to the control focus theory.

Referring to FIG. 2, the consumer information collecting unit 110 pre-sets consumer tendency information to classify the consumer's disposition according to the control focus theory together with basic personal information such as name, gender, birthday, and occupation from the consumer. Can be input. The consumer propensity information may be expressed as scores (1 to 5) input from the consumer for each of six preset questions (survey items). In this way, the consumer's disposition may be classified into an improved focus or a preventive focus based on the input score.

Referring back to FIG. 1, the consumer information collecting unit 110 collects a web shopping log indicating an action performed by a consumer to purchase a product on an online site. In this case, the products sold on the online site are not limited to the goods, and various types such as services and contents are possible. In addition, there is no particular limitation on the way of selling products on the online site, and an intermediary selling method such as an open market is possible.

The collected web shopping log refers to data accumulated in a transaction based on a time when a click event of a consumer occurs in an online site for purchasing a product. The consumer information collecting unit 110 includes a consumer's unique identification information (consumer ID), a unique identification information (product ID) of a product searched by the consumer, and a page of an online site including the product searched by the consumer. Unique shopping information (review ID) of the clicked product review, and when the consumer accesses the page can be collected in the web shopping log.

Table 1 below shows an example of the web shopping log collected by the consumer information collecting unit 110.

Consumer ID Product ID Review ID Page
Connection point Behavior information Consumer behavior types 123 89 256 2012-07-27
18:59 PR_SUMMARY_ <Click Review> See online reviews 125
35
empty
2012-08-03
11:40 BOTTON_ <cart> Purchase intent

Referring to Table 1, it can be seen that the consumer 123 accesses the review 256 of the product 89 at 18:59 on July 27, 2012. This behavior of consumer 123 corresponds to the type of consumer behavior of the online review reference. Similarly, Consumer 125 put a commodity 35 in the shopping cart at 11:40 on August 3, 2012. This behavior of consumer 125 corresponds to the type of consumer behavior of purchase intention.

The information in Table 1 may be extracted by analyzing the URL value of the link clicked by the consumer and the parameter value delivered by the GET method or the parameter value delivered by using JavaScript. Here, the consumer behavior type is information that can be extracted from the web shopping log through research in the field of consumer behavior. The type of product navigation related to the search for information about a specific product and the alternative navigation type related to the total number of product pages accessed by the consumer Can be classified as.

Product search types include types of online review references, purchase intentions, and integrations. An online review reference is a type of action related to a product-related review or to a general review related to the service, such as shipping, cancellation, or refund, that includes clicking on a rating such as highly recommended, recommended, moderate, or not recommended for each review. This is the case. In addition, a purchase intention is a type of action related to purchasing a specific product, which corresponds to an action of putting a specific product into a shopping cart, clicking an item such as buying or wishing to steam. An integration type is a type of action that involves the number of clicks on a particular product page on an online site or the amount of time it takes to click a specific piece of information.

The search log calculator 120 calculates a search log indicating whether the consumer searches for preset product related information. In order to calculate the search log, the search log calculator 120 first calculates a click event log. The click event log is a click-by-click representation of the consumer's actions on a page of the online site that contains the product the consumer retrieved based on the collected web shopping log. In detail, the click event log may be calculated by Equation 1 below.

로 구성되며 아래 첨자는 발생 시점의 시간 순서를 의미한다. 즉,

은 특정 소비자 c가 검색한 특정 상품을 포함하고 있는 온라인 사이트의 특정 페이지 p에서 발생한 m번째 클릭 이벤트 로그를 의미한다. Here, T ^{c , p} denotes a click event log generated on a specific page p of the online site including a specific product searched by a specific consumer c. T ^{c , p} total m click events

The subscript means the time sequence at the time of occurrence. In other words,

Denotes the mth click event log generated on a specific page p of an online site containing a specific product searched by a specific consumer c.

Thereafter, the search log calculator 120 calculates a search log based on the calculated click event log. As described above, the search log indicates whether the consumer searches for preset product related information.

Table 2 below shows the product related information and the search log.

consumer
Behavior type Product related information Click count Clicked Time



online
See review

Product reviews ○ ○ X Product review rating
(Highly recommended, recommended, normal, not recommended, etc.) ○ ○ X General review ○ ○ X General review rating
(Highly recommended, recommended, normal, not recommended, etc.) ○ ○ X

Purchase intent
shopping basket X ○ ○ buying X ○ ○ Wish X ○ ○ Add to Cart X ○ X integrated Inside product page ○ X ○

Referring to Table 2, product-related information includes information about the consumer's opinion about a specific product being sold on an online site (see an online review), information indicating whether the consumer has purchased a specific product (purchasing), It may include information (shopping basket, steaming) indicating whether or not you want to purchase. Specifically, the product-related information includes general reviews related to the service, such as product-related reviews, shipping, cancellations, and refunds, ratings such as highly recommended, recommended, moderate, and not recommended for each review, add to cart, buy, wish The following items may apply. It can be seen that such product-related information corresponds to the consumer behavior type described above.

Accordingly, the search log includes the number of clicks on product-related information, whether or not the product-related information is clicked, and the time it takes to click on the product-related information from the time of accessing the page of the online site containing the product-related information. Can be. In Table 2, o indicates a suitable item to calculate from the search log.

Specifically, the number of clicks on the product related information refers to the number of transactions in which the specific consumer c clicks on the specific product related information x on the specific product page p. Whether or not the product-related information is clicked means whether or not a transaction in which a specific consumer c clicks on a specific product-related information x on a specific product page p. Also, the time required refers to the time taken for a specific consumer c to access a specific product page p and click on the specific product related information x included in the specific product page p.

The behavior log calculator 130 calculates a behavior log indicating whether the consumer has performed an action corresponding to the above-described consumer behavior type based on the search log. The behavior log may be calculated for a specific consumer or for a specific consumer and a specific product page. That is, as described above, since the search log is calculated for each specific product page, the behavior log may be calculated for each specific consumer by obtaining a total, average value, maximum value, minimum value, and standard deviation for each click count or time required.

The activity index calculator 140 calculates an activity index indicating a degree to which a consumer searches for information necessary to purchase a product. Activity index can be calculated by the following equation (2).

는 특정 소비자 c가 특정 상품 페이지 p에서 상품 관련 정보를 활발히 검색하였는가를 판단할 수 있는 결정 함수(Decision Function)로 0 또는 1의 값을 갖는다. 그리고 q는 특정 소비자 c가 검색한 상품 중 쾌락재(Hedonic Goods)에 속하는 상품 페이지의 수를 의미한다. here,

Is a decision function for determining whether a specific consumer c actively searches for product related information on a specific product page p, and has a value of 0 or 1. Q denotes the number of product pages belonging to Hedonic Goods among the products searched by a specific consumer c.

The theory of consumer behavior is generally described by dividing the product into Hedonic Goods and Utility Goods. Pleasure goods are products that provide emotions such as pleasure, joy and positive excitement, such as flowers, clothing, music, sports cars, luxury watches, and chocolate. In addition, the utility material means a product in which the functional elements of the product are important, such as a microwave oven, a minivan, a home security system, and a personal computer. Whether or not a particular product is a pleasure or utility is determined by the consumer's theory of consumer behavior in which he or she perceives a relatively more of a hedonic or a practical one.

를 생성하기 위해 사전에 설정한 기준으로 행동 로그를 분류할 수 있는 군집화 알고리즘을 이용할 수 있다. 구체적으로, K-평균 군집화 알고리즘을 이용하여 특정 소비자 및 특정 상품 페이지별로 산출한 행동 로그를 두 개의 군집으로 분류한다. 그 후 앞서 설명한 소비자 행동 유형에 해당하는 온라인 리뷰 참조 및 통합 유형과 관련된 수치가 평균적으로 높은 군집을 활발한 군집으로 정의하고 나머지 군집을 활발하지 못한 비활동성 군집으로 정의할 수 있다. In addition, the activity index calculation unit 140

A clustering algorithm can be used to classify the behavior log based on a predetermined criterion to generate the. Specifically, the behavior log calculated for each specific consumer and specific product page is classified into two clusters by using the K-means clustering algorithm. You can then define an average cluster with a high number of online reviews references and integration types that correspond to the types of consumer behavior described above as active clusters, and the remaining clusters as inactive clusters.

The classification model generator 150 generates a rule for classifying a consumer's propensity to access the online site based on the calculated behavior log and activity index and the consumer propensity information received from the consumer. Specifically, the classification model generator 150 combines the calculated behavior log and activity index to calculate CV (Cross Validation) data. However, in this combination, the collision between the semantically overlapping variables decreases the accuracy. Therefore, it is necessary to select important variables that have a great influence on the propensity classification.

In this case, the classification model generator 150 may select important variables based on two main conditions. First, the importance level for each variable is extracted by calculating the p-value based on the F-Statistic. The variables are classified according to the above-described consumer behavior type and the descending order is performed based on the importance. Then, the variables with the largest values within each consumer behavior type are selected as important variables to eliminate variables with duplicate meanings.

3 is a table showing an example of a process for selecting important variables.

Referring to FIG. 3, Alternative ProductPageCnt corresponding to alternative search type, WishListClickBin_Min corresponding to purchase intention, and Exploration ActivityIndex corresponding to online review reference and integration were selected as important variables.

The classification model generation unit 150 generates a rule that can classify the propensity of consumers to access the online site based on the classified consumer preferences based on the selected important variables and the consumer propensity information input from the consumers. . In this case, the classification model generator 150 may classify the consumer's disposition based on the consumer disposition information.

FIG. 4 is a graph showing the relationship between consumer preferences classified based on important variables selected through the process of FIG. 3 and consumer orientation information.

Referring to FIG. 4, (a) shows the distribution of consumer dispositions according to the number of alternatives searched by the consumer, (b) shows the distribution of consumer dispositions according to whether the product is put in a shopping cart or steamed, and (c) ) Shows the distribution of consumer disposition according to activity index.

In detail, the classification model generator 150 uses machine learning algorithms to generate a rule. As a machine learning algorithm, algorithms such as Decision Tree, Naive Bayesian, Support Vector Machine (SVM), and Neural Network (Multilayer Perceptron) may be used. Decision tree algorithms are particularly desirable because they facilitate consumer theory of behavior and cross checking.

FIG. 5 is a diagram illustrating a decision tree generated based on an important variable selected through the process of FIG. 3.

Referring to FIG. 5, it can be seen that consumers who have carried out shopping carts or steamed shops with respect to all products searched based on a minimum value of whether shopping carts or shoppers are likely to be preventive. We also found that consumers with a low history of shopping carts or shopping carts are more likely to have an improvement if their activity index is less than or equal to -1, or they are more likely to have an improvement if the number of alternatives searched is greater than 5. Can be.

Referring back to FIG. 1, the consumer disposition classification unit 160 classifies the consumer's disposition based on the behavior log and activity index calculated for the consumer accessing the online site and the generated rule. In this case, if the above-described rule has already been generated, the consumer information collecting unit 110 does not receive the consumer propensity information for the consumer separately from the consumer, and the consumer disposition classification unit 160 calculates the behavior log and activity index calculated for the consumer. And classify the consumer's disposition based on the generated rules.

6 is a flowchart illustrating a preferred embodiment of the method for classifying the adjusted focus tendency of a consumer according to the present invention.

The consumer information collecting unit 110 receives the consumer tendency information that can classify the consumer's propensity according to the pre-set consumer's propensity classification from the consumer who accesses the online site, and performs the consumer's purchase on the online site. The web shopping log indicating the behavior is collected (S610). The web shopping log refers to data accumulated by a transaction based on a time when a click event of a consumer occurs in an online site for purchasing a product.

The search log calculator 120 calculates a search log indicating whether the consumer searches for preset product related information (S620). In order to calculate the search log, the search log calculator 120 first calculates a click event log. The click event log is a click-by-click representation of the consumer's actions on a page of the online site that contains the product the consumer retrieved based on the collected web shopping log.

In addition, the behavior log calculator 130 calculates a behavior log indicating whether the consumer performs the behavior corresponding to the above-described consumer behavior type based on the search log (S620).

Next, the activity index calculation unit 140 calculates an activity index indicating the degree to which the consumer searches for information necessary to purchase a product (S630). To this end, the activity index calculation unit 140 may use a clustering algorithm that can classify the behavior log based on a preset criterion.

Thereafter, the classification model generator 150 generates a rule for classifying the propensity of the consumer to access the online site based on the calculated behavior log and activity index and the consumer propensity information input from the consumer (S640). In detail, the classification model generator 150 may generate a rule using machine learning algorithms such as decision trees, naive Bayesian, SVM, and neural network algorithms based on consumer inclinations and important variables classified through consumer inclination information. Can be.

Finally, the consumer disposition classification unit 160 classifies the consumer's disposition based on the behavior log and activity index calculated for the consumer accessing the online site and the generated rule (S650).

The present invention can also be embodied as computer-readable codes on a computer-readable recording medium. A computer-readable recording medium includes all kinds of recording apparatuses in which data that can be read by a computer system is stored. Examples of the computer-readable recording medium include a ROM, a RAM, a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like, and may be implemented in the form of a carrier wave (for example, transmission via the Internet) . The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is clearly understood that the same is by way of illustration and example only and is not to be taken by way of limitation in the embodiment in which said invention is directed. It will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the scope of the appended claims.

Claims (19)

Web shopping that indicates an action performed by the consumer to purchase a product on the online site by receiving consumer propensity information for classifying the consumer's propensity according to a pre-set consumer's propensity classification from a consumer who accesses the online site A consumer information collecting unit for collecting logs;
A search log calculator configured to calculate a search log indicating whether the consumer searches for preset product related information;
A behavior log calculator configured to calculate a behavior log indicating whether the consumer performs a behavior corresponding to a preset consumer behavior type in response to the product related information based on the search log;
An activity index calculator for calculating an activity index indicating a degree to which the consumer searches for information necessary to purchase a product;
A classification model generator for generating a rule for classifying a consumer's disposition to access the online site based on the calculated behavior log and activity index and consumer disposition information received from the consumer; And
And a consumer propensity classifier that classifies the consumer's propensity based on the activity log and activity index calculated for the consumer accessing the online site and the generated rule. The method of claim 1,
The consumer information collection unit may be unique to the product review clicked by the consumer on a page of an online site including the unique identification information of the consumer, the unique identification information of the product searched by the consumer, and the product searched by the consumer. And at least one of identification information and a time point at which the consumer accesses the page, as a web shopping log for the consumer. The method of claim 1,
The search log calculation unit calculates a click event log indicating, by click, an action performed by the consumer on a page of an online site including the product searched by the consumer based on the collected web shopping log, and the calculated click And a search log based on an event log. The method of claim 3, wherein
For each product that sells the product-related information on the online site, a consumer's opinion about the product, information indicating whether the consumer has purchased the product, and information indicating whether the consumer wants to purchase the product Apparatus classification apparatus of the consumer, characterized in that set to at least one of. 5. The method of claim 4,
The search log calculation unit takes the number of clicks on the product related information, whether the user clicks on the product related information, and the time required to click on the product related information from the time of accessing the page of the online site including the product related information. At least one of the time to calculate the search log characterized in that the consumer tends to classify. 6. The method of claim 5,
And the behavior log calculator calculates at least one of the total number, the average value, the maximum value, the minimum value, and the standard deviation of the number of clicks or the time required as the behavior log for the consumer. The method of claim 1,
And the activity index calculator calculates the activity index using a clustering algorithm capable of classifying the behavior log based on a preset criterion. The method of claim 1,
The classification model generator may classify the consumer's propensity based on the consumer's propensity classified through the consumer propensity information input from the consumer accessing the online site, the behavior log calculated for the consumer, and the activity index. Apparatus classifying consumer's disposition, characterized in that for generating a rule using a machine learning algorithm (Machine Learning Algorithms). The method of claim 1,
The consumer information collecting unit does not receive the consumer disposition information for the consumer from the consumer accessing the online site,
The consumer disposition classifier classifies the consumer's disposition according to the behavior log and activity index calculated for the consumer and the generated rule. Web shopping that indicates an action performed by the consumer to purchase a product on the online site by receiving consumer propensity information for classifying the consumer's propensity according to a pre-set consumer's propensity classification from a consumer who accesses the online site Collecting consumer information to collect logs;
A search log calculating step of calculating a search log indicating whether the consumer searches for preset product related information;
A behavior log calculation step of calculating, based on the search log, a behavior log indicating whether the consumer has performed an action corresponding to a preset consumer behavior type in response to the product related information;
An activity index calculation step of calculating an activity index indicating a degree to which the consumer searches for information necessary for purchasing a product;
A classification model generation step of generating a rule for classifying a consumer's propensity to access the online site based on the calculated behavior log and activity index and consumer propensity information received from the consumer; And
And a consumer disposition classification step of classifying the disposition of the consumer based on the activity log and activity index calculated for the consumer accessing the online site and the generated rule. The method of claim 10,
In the collecting of the consumer information, the product review clicked by the consumer on a page of an online site including the unique identification information of the consumer, the unique identification information of the product searched by the consumer, and the product searched by the consumer And at least one of unique identification information and a time point at which the consumer accesses the page, as a web shopping log for the consumer. The method of claim 10,
In the search log calculation step, a click event log indicating an action performed by the consumer on a page of an online site including a product searched by the consumer based on the collected web shopping logs is calculated, and the calculation is performed. And calculating the search log based on the click event log. 13. The method of claim 12,
For each product that sells the product-related information on the online site, a consumer's opinion about the product, information indicating whether the consumer has purchased the product, and information indicating whether the consumer wants to purchase the product The consumer's disposition classification method, characterized in that set to at least one of. 14. The method of claim 13,
In the search log calculation step, the number of clicks on the product-related information, whether or not the product-related information is clicked, from the time of accessing the page of the online site including the product-related information until the click of the product-related information And calculating at least one of the required time periods as the search log. 15. The method of claim 14,
In the behavior log calculation step, at least one of the total number, the average value, the maximum value, the minimum value, and the standard deviation of the number of clicks or the required time is calculated as the behavior log of the consumer. Way. The method of claim 10,
In the activity index calculation step, the consumer's propensity classification method, characterized in that for calculating the activity index using a clustering algorithm that can classify the behavior log on a predetermined basis. The method of claim 10,
In the classification model generation step, the propensity of the consumer is classified based on the propensity of the consumer classified through the consumer propensity information input from the consumer accessing the online site, the behavior log calculated for the consumer, and the activity index. Method for classifying consumer's disposition, characterized in that it generates a possible rule using a machine learning algorithm (Machine Learning Algorithms). The method of claim 10,
In the consumer information collection step, the consumer disposition information for the consumer is not received from the consumer accessing the online site,
In the consumer disposition classification step, the disposition of the consumer characterized in that the disposition of the consumer based on the behavior log and activity index calculated for the consumer and the generated rule. A computer-readable recording medium having recorded thereon a program for executing the method of classifying consumer's disposition according to any one of claims 10 to 18.

Images