KR20080083248A

KR20080083248A - Online item recommendation system built upon users' "like or dislike"choices - based on statistical similarity between users' choices

Info

Publication number: KR20080083248A
Application number: KR1020080083652A
Authority: KR
Inventors: 이종운
Original assignee: 이종운
Priority date: 2008-08-25
Filing date: 2008-08-25
Publication date: 2008-09-17
Also published as: WO2010024535A2; WO2010024535A3

Abstract

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. The item recommendation system according to the present invention includes a DB in which a corresponding "user ID", "item title", and "good / dislike evaluation value" are recorded when a user (i) evaluates the item as "good" or "dislike"; A sympathetic probability function for calculating similarity of good / bad ratings between users from the DB; A recommendation item chart that automatically sorts users who have rated items similar to me according to the sympathy probability and items that they said are good; Empathy user search function to find users who have rated the same as the particular item (s) from the DB: A searched empathetic user chart consisting of a recommendation item chart to manually sort items.

Description

Online item recommendation system built upon users '“like or dislike” choices-based on statistical similarity between users' choices}

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, which requires the identification of 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.

[Solution solution and effect]

The present invention is a method in which a user searches for other users who have evaluated the most similar to their own and recommends other items which are said to be good as long as the user evaluates the like / dislike of items on the Internet. Users do not need to disclose their personal information, just by evaluating items that are good / dislike to find other users who have the statistically similar tastes as theirs, and receive the items with the highest preferences from them. The system recommends numerous items to the tastes of many people without invading privacy, thereby increasing the desire, participation and satisfaction of Internet users.

The item recommendation system according to the present invention includes a DB in which a corresponding "user ID", "item title", and "good / dislike evaluation value" are recorded when a user (i) evaluates an item as "good" or "dislike"; A sympathetic probability function that calculates similarity of good / bad ratings between users from the DB; A recommendation item chart that automatically sorts users who have rated items similar to me according to the sympathy probability and items that they said are good; A sympathetic user search function for finding users who have made the same evaluation as the specific item (s) from the DB; It consists of a chart of recommended items that manually sorts the items found by empathy users.

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

1 is a block diagram of a system for recommending an item closest to a user's taste according to a probability of empathy. First, a user 1 connects to an Internet website to which the recommendation system is applied. (S110)

The user 1 who accesses the website evaluates the good / dislike of the item list 2A randomly exposed on the website 2. (S120) The item list can be anything, such as books, music, movies, advertisements, figures, pictures, abstract concepts, etc., and the exposure method can be any form such as a chart or a popup.

The list of items that the user liked / disliked is stored in the user-item DB 3A of the server 3 so that each user has rated which item, which user has given a good rating for each item, You can see who disliked you. In the DB 3A, " User ID " " Item Title " " Good / Like Evaluation Value (for Item) "

When the user 1 makes a good / dislike evaluation of an item, the users who evaluate the item most similar to the self are automatically searched by the empathy probability function 3B of the present invention.

The empathy probability function (3B) is a calculation that tells users how many of the same evaluations are made on the same item, and automatically selects other users who have made the most similar choices to a specific user, and selects those whose tastes are closest to the specific user. Sift The empathy probability function 3B is calculated in three steps.

1) Check if the user 1 has rated / disliked 5 or more items. (S130) In order to increase the reliability of the probability, a minimum number of samples must be secured. Since the sympathy probability is a task of matching the user's response to the item, a user who evaluates at least five items must be secured. If the user 1 does not like or dislike 5 or more items, the probability of empathy is not calculated.

2) If the user 1 has rated five or more items, search for other users who have liked / disliked at least five of the items (the five or more) that the user 1 rated. (S140) If the user (1) has evaluated seven items, there must be at least one other user who has evaluated five items, and if the user (1) has not rated any one of the five items among the seven items, the user (1) Empathy probability is excluded from the calculation. Also, for example, if user (1) rated seven items, and user A only evaluated four of them, user (1) and user A only evaluated four items, not five, so user A could It is excluded from the calculation of empathy probability of (1).

3) The empathy probability value for the user 1 is obtained for the users found in step 2. (S150) The sympathy probability is a ratio that calculates how many same ratings each user has on the same number of items as the user 1. Empathy probability = the number of items that the user (1) and the counterpart user gave the same evaluation ÷ the number of items that the user (1) rated and also the counterpart user rated

For example, user 1 rated seven items, user A rated five of seven items evaluated by user 1, and user 1 rated two of these five items. If the following evaluation is made, the probability of the user A's empathy with respect to the user 1 is 2 ÷ 5 = 40%.

A user list 3C is created in which the users retrieved in step 2 are sorted in ascending probability values calculated in step 3. (S160)

Finally, create an item chart 2B that sorts the items that the sorted users said good. (S170) The item chart sequentially shows the items that the user who has the highest empathy probability is good. At this time, the items that the user 1 has already evaluated are shown.

Example of an auto recommended item chart (2C)

By viewing the item chart 2B, the user 1 can find the items that he or she did not know, most likely to suit his or her taste. If you don't like the item chart, you can increase the number of target users and increase the accuracy of your empathy probability by rating more / more like more items. (S180)

2 is a system block diagram in which a user is recommended an item through users who have given the same evaluation as a specific item. First, the user 1 accesses an internet website to which the recommendation system is applied. (S210)

The user 1 evaluates the like / dislike for a number of items shown in the item evaluation module 2A of the website 2. The item may be anything, such as a book, music, movie, advertisement, person, picture, abstract individual, etc. The item evaluation module may be in any form such as a chart or a popup.

If the items automatically sorted by the sympathy probability do not like the user, the user 1 selects only the specific item (s) among the items which he / she likes / dislikes and searches for those who have given the same rating as this item. Can be.

For example, a user who enjoyed the book "DaVinci Code" was really fun, but the movie "Harry Potter" wasn't really fun. "Harry Potter" can only filter out users who have disliked it.

The user 1 selects an item that has been evaluated by those who have evaluated the same item among the evaluated items and designates it as a sympathetic item in the sympathetic item designation module 2D. (S230)

When a sympathetic item is designated, the server 3 searches for a sympathetic user list 3E having the same empathy as the user 1 on the sympathetic item search 3D function. (S240)

Empathic users are searched by and operations when two or more empathy items are specified. That is, when two or more empathy items are designated, both items are searched for users who gave the same rating as themselves.

When the list of sympathetic users is searched, items that the sympathetic users said to be good are searched together. Items that have already been evaluated by the user 1 are excluded from the searched items, and the items are sorted in order of receiving the most favorable ratings from the sympathetic users and exposed in the form of a recommendation chart 2E. (S250)

While the recommendation item chart based on the sympathy probability is automatically provided without a user's command, the recommendation item chart by the sympathetic user is a passive chart system in which the user must select an item directly.

Manual Recommended Item Chart (2E) Example

By evaluating more items and specifying more empathy items, the user 1 can reduce the number of empathy users and recommended items and recommend items that are more suitable for their tastes. (S260)

1 is a block diagram of a system for suggesting an item closest to a user's preference according to a sympathy probability

2 is a block diagram of a system in which a user is recommended an item through users who have given the same rating to a specific item (s).

3 is a flowchart of a system for recommending an item closest to a user's taste according to the sympathy probability

4 is a flow diagram of a system in which a user is recommended an item through users who have given the same rating to a specific item (s)

Explanation of symbols for the main parts of the drawings

1: user client

2: website

2A: Randomized Item List 2B: Auto-Sorted Featured Item Chart

2D: Empathy Item Assignment Module 2E: Manually Ordered Recommended Item Chart

3: server

3A: User-Item DB 3B: Empathy Probability Function 3C: List of Users extracted with Empathy Probability

3D: Empathy Users Search 3E: Empathy Users List

Claims

If users rate like / dislike a large number of items, a DB that records "user ID", "item title", and "good / like evaluation value (for items)" is recorded,

A sympathetic probability function that calculates similarity of good / bad ratings between users from the DB,

A list of users who search-sort the users who have rated the most similar to you according to the value of the empathy function,

Personalized-recommended system, characterized in that consisting of a recommendation item chart that automatically sorts the items that the sorted users said good according to the sympathy probability value order.

The method of claim 1 wherein the empathy probability function is

In a 1: 1 relationship between users, we calculated the ratio of how many items were given the same rating to each other,

1) confirming whether a particular user (I) rated good or dislike for 5 or more items;

2) if it is determined in step 1) that I have rated at least 5 items, searching for other users who have rated good / dislike of at least 5 of the items I have rated;

3) The ratio of the number of items that the users searched in step 2) gave the same evaluation to the same number of items and the number of items that the other user and the other user rated the same ÷ me and the other user also evaluated) Calculation step

User search and matching process, characterized in that consisting of.

The recommended item chart of claim 1, wherein

Auto-recommendation function in the form of a chart, characterized in that the items found by the users found by the sympathy probability is ranked in order of high sympathy probability for each user, except for items that have already been evaluated by the sorted items.

Empathy user search function that finds users who rated a particular user (me) like / dislike a number of items, and then rated all of the same on one or many items,

Personalized-recommended system, characterized in that consisting of the recommended item chart to sort the items that the users searched by the empathic user search function.

The recommended item chart of claim 4, wherein

Manual recommendation function in the form of a chart characterized in that the items found by the sympathetic users found through the sympathetic user search is sorted in order of receiving the highest ratings from the sympathetic users, except for items that I have already rated.