KR20080087076A - Online recommendation system built upon users' "like or dislike" choices - based on user classification - Google Patents

Abstract

An Internet system for receiving item recommendation based on statistical similarity of user evaluation by making a user choice like or dislike is provided to recommend an item statistically having the highest like probability to a user by making the users choose like or dislike of the item without opening any personal information. A user type-suitability database(3A) classifies users(1A) into each type and stores user suitability for the type. An item type preference database(3B) stores the type suitability of the classified user(1B) as a like/dislike score when the classified user evaluates like/dislike of the item. An index item evaluating module(2C) adjusts the type suitability of the user by appointing and using the item having remarkably high or low preference in the predetermined type. A recommended item chart(2D) recommends the items having the highest preference probability in the type corresponding to the user. The index item evaluating module automatically appoints the item having the higher type suitability than 80 percents and the item having the lower type suitability than 20 percents as the index item of the type, and varies the type suitability of the user evaluating the index item.

Description

Online recommendation system built upon users' “like or dislike” choices-based on user classification}

The present invention relates to a personalized recommendation system on the Internet that recommends items that are closest to the tastes of individual users of a website. Unlike the existing personalization service that had to analyze a member's age, gender, website visit history, etc. It is characterized by the fact that only by voluntary good / dislike of item evaluation.

If you use the Internet to find what you're looking for, you'll often see too many items or things you don't like. That's why existing websites offer Internet personalization and recommendation services. The current Internet personalization service records personal information such as age, gender, and place of residence, and personal information such as search, clicks, and purchased items, and the recommendation service is made based on recommendation, empathy, and scrap number by a large number. .

Current Internet personalization services invade privacy by recording various personal information, and recommendation services are often based on voting by the majority and do not meet personal preferences.

The present invention is a method of recommending items with the highest preference in the same type as the user when the user determines his or her type on the Internet and evaluates like / dislike items. Users do not need to disclose their personal information at all and decide their type and evaluate items that are good / disliked by the item. The system recommends numerous items to the tastes of millions of people without invading privacy, thereby increasing the desire, participation and satisfaction of Internet users.

The present invention relates to a system for recommending an item closest to a user's taste based on the rating of a user who likes / dislikes a plurality of items. An item recommendation system according to the present invention comprises: a user type-fit DB for classifying users into a predetermined type and storing the user's fitness for the type; A classification of the item type preference DB for recording the type suitability of the user as a preference / rejection score for each type of the item when the classified user is satisfied with / disliked the item; An indicator item evaluation module for separately designating items with high or low preference in a specific type, and adjusting the type suitability of the user with the item; It consists of an item chart that recommends the items that are most likely to be preferred for each user type.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a block diagram of a system for recommending an item closest to a user's taste according to a type classification. First, a user 1A connects to an Internet website to which the present recommendation system is applied. (S10)

The user 1A accessing the website participates in the type classification 2A prepared in advance, and determines his or her type. (S20) The type classification can be done by the user selecting the type directly or through a test like the MBTI psychological test. Types can also be categorized based on population distribution, such as gender, age, or education level, and set up as groups of people who respond differently to specific objects such as movies, music, and advertisements. The same type should respond similarly to the same object, while different types should respond differently to the same object.

A user who participates in the type classification is registered as a classified user 1B, and is given a type and type suitability score, which is recorded in the user type-conformance DB 3A. (S30) The fitness score is a score indicating how suitable the user is for the type. At the time of class classification, only the basic score is given, and then it is changed by a good / dislike evaluation of the indicator item.

Example of data written to user type-conformance DB (3A)

User type-conformance DB 3A can record more than one type for a user. This is because a goodness-of-fit 10 for type A and a goodness-of-fit 5 for type B can be shown by one or more types. The user type-conformance database 3A records the fitness for each type and updates the variation in each fitness.

The categorized user 1B makes good / like disapproval on several items in the item evaluation module 2B of the website. Users 1A not classified in the type S40 may not like or dislike the item in the item evaluation module. The categorized users 1B are given a rating that is good / dislike for the items exposed to the item rating module 2B. Items can be anything from books, music, movies, advertisements, figures, drawings, abstract concepts, and more.

An item that has been evaluated as good or dislike of the classified user 1B is accumulated in the type suitability score of the user and recorded in the item-specific preference DB 3B. (S50) When the evaluation of various types of users is accumulated, it is possible to determine from which type the items are more preferred / rejected from the preference DB 3B for each item type.

For example, if user A of type A, goodness of fit 10 evaluates that item A is good, item A has a score of 10 points of preference. Subsequently, if a user B of type A and goodness of fit 10 evaluates item A as dislike, item A has 10 rejection points recorded in type A. In this way, the preference (good) and rejection (dislike) scores of a plurality of users are summed by type, and the preference score is calculated by calculating the ratio of the preference scores among the total scores of the combined scores and preference scores of each type. (Preferences = Preferred Scores ÷ (Preferred Scores + Rejected Scores))

Example of data recorded in the preference DB (3B) by item type

Some of the items may have very high or low preferences for a particular type. If a preference is above 80% or less than 20% in a particular type, this item is designated as an indicator item of that type. (S60) If a preference type of 80% or more is recorded in a specific type, a preference indicator item of that type is designated as a rejection indicator item of the type if a preference of 20% or less is recorded in a specific type. An index item can only be specified if there are more than 10 Good / Like ratings for that type.

If the sorted users 1B do not like the indicator items in the indicator item evaluation module 2C (S70), their type suitability scores are changed. (S80) The user's fitness score changed by the evaluation of the indicator item is recorded in the user type-conformance database 3A, and the fitness score is re-granted to the classified users. (S30)

For example, if User A classified as Type A dislikes an item that other users of the same type are overwhelmingly good, then User A has made an inappropriate rating for Type A, and other users of the same type are saying that he or she is good at If so, the evaluation was appropriate. On the contrary, if other users of the same type say they prefer an item that they overwhelmingly dislike, User A has made an inappropriate rating for Type A. At this time, the criterion for judging whether the evaluation is suitable or not is an index item.

Assume that item A is a preference indicator item that shows 82% preference in type A. If type A's user A disagrees with item A, then user A has made an inappropriate choice for type A, and therefore user A's suitability for type A is diminished. If Type A's User B evaluates to be good for Item A, then User B has made a suitable choice for Type A, and therefore User B's fitness for Type A is raised.

Assume that item A is a reject indicator item with a 6% preference in type B. If user B of type B evaluates to be good for item A, user C has made an inadequate choice for type B, thus reducing user C's suitability for type B. If type B's user D says he dislikes item A, then user D has made a suitable choice for type B, and so user D's suitability for type B increases.

The fitness scores fluctuated by the index item evaluation are not fixed but may vary depending on the basic fitness score, the number of index items, and the amount of activity of website users.

The user fitness score changed by the indicator item is accumulated as a preference / rejection score at the next item evaluation (S40). (S50) The good / dislike evaluation of the user whose quality is reduced has a relatively small influence on the preference of the item, and the good / dislike evaluation of the user who has increased the fit has a relatively large influence on the preference of the item. If the goodness-of-fit score continues to drop to 0, this person's fitness for that type is 0, which is good / dislike for the item, but it is not recorded in the item-specific preference DB (3B), which does not affect the item-specific preferences. .

Based on the data accumulated in the preference DB (3B) for each item type, the classified users (1B) can view the items that are suitable for them by viewing the recommended item chart (2D) that arranges the items with the highest preference in their type. have. (S90)

If the items in the recommendation item chart 2D are unsatisfactory, the user may rejoin the type classification 2A to change his or her type or select the type with the recommendation item chart that is most suitable for him. (S100)

1 is a block diagram of a system for recommending an item closest to a user's taste according to a user type classification.

2 is a flowchart of a system for recommending an item closest to a user's taste according to a user type classification.

Explanation of symbols for the main parts of the drawings

1: user client

1A: User (unclassified)

1B: Users classified (by type classification)

2: website

2A: Type Classification

2B: Item Evaluation Module

2C: Indicator Item Evaluation Module

2D: (by Type) Recommended Item Chart

3: server

3A: User Type-Fitness DB

3B: Affinity DB by Item Type

Claims (6)

A user type-fit database that categorizes users into a given type and stores the user's goodness-of-fit for that type; When a classified user evaluates a good / dislike of an item, a preference DB for each item type, which records the user's type suitability as the item's type preference / rejection score, An indicator item rating module that specifies items with high or low preference for a particular type, and controls the type fit of the user with these items, An Internet recommendation system comprising an item chart recommending items that are most likely to be preferred in the type of user. The method of claim 1 wherein the user type-conformance DB, DB, characterized in that one or more types can be recorded in one user, and a change in the fitness value for each type can be recorded. The method of claim 1, wherein the preference DB by item type is Distinguishing the like / dislike rating of the user by the type of the user is recorded as a preference / rejection score for each type, and this preference / rejection score is changed as the fitness value for the user's type changes. DB. The method of claim 1, the indicator item evaluation module, A device type suitability control apparatus characterized by automatically assigning an item having a preference of 80% or more and an item having a preference of 20% or less as an indicator item of the type, and changing the type suitability of a user who evaluates the indicator item. The indicator item of claim 4, If a particular type has a preference of 80% or higher, the user's rating of that type is good, and that type's fitness increases.If you dislike it, the type's fitness decreases. Criteria for adjusting user type suitability, characterized by an increase in type suitability, and a good rating decreases. The item chart of claim 1, wherein A chart-like recommendation feature that sorts items that have a high preference (preferred score ÷ (preferred score + reject score)) in each user type in order of preference.

Images