運用小眾愛好做好友推薦的社交匹配系統

Abstract

好友推薦是社會化網站的重要應用之一。好友推薦系統（Social matching system）幫助用戶找到他們感興趣的人。然而對於新用戶，由於沒有新用戶現有的好友、行為記錄，無法為新用戶定製一份個性化的好友推薦。這時候推薦系統必須提供一個有效的偏好獲取過程，給新用戶表達他們的偏愛和喜好，然後給新用戶做推薦。傳統的獲取用戶偏好的方式，是讓用戶給一組熱門物品評分。然而有研究發現，對於共同擁有小眾特徵、愛好的同類，用戶通常會更感興趣。基於這個研究的發現，我們設計了一個運用小眾愛好做好友推薦的系統，想要驗證在給用戶做好友推薦時，重視小眾愛好所得到的推薦結果會讓用戶更滿意。在受測者實驗中，運用小眾特征的系統，和傳統方式的系統相比，取得了更好的結果，用戶的滿意度更高，實驗中有不少有趣的發現，可以給設計好友推薦系統的人提供參考。由於讓受測者給電影評分的數量很少，類似於冷啟動問題的場景，因此這個好友推薦系統也可以應用在冷啟動問題。

Friend recommendation is one of the important applications of socialization website. A friend referral system helps users find people they are interested in. However, when a buddy recommendation is made to a new user, there is no record of the user's existing friends and users, a personalized recommendation for a new user is recommended, and the user is satisfied with the recommendation result, and is willing to continue using the system. Must provide an efficient, effective process, to give new users to express their preferences and preferences. The traditional way of capturing user preferences is to allow users to rate a group of popular items. However, there are user research that, for the common characteristics of a small minority, like the kind of user is usually more interested. Based on the findings of this user study, we designed a system using a small minority hobby buddy recommended to do, want to verify, to the user to do a friend recommendation, attention to small hobbies, get the recommendation results will allow users more satisfied. In the experiment, the use of minority characteristics of the system, compared with the traditional way of the system, and achieved better results, higher user satisfaction, the experiment there are many interesting discoveries, you can recommend to the design of friends System of people to provide reference. Due to the limited number of subjects allowed to rate the movie, similar to the cold start scenario, a cold start related technique was applied. This buddy recommendation system can also be applied to cold start problems.