Novel personal and group-based trust models in collaborative filtering for document recommendation

Abstract

Collaborative filtering (CF) recommender systems have been used in various application domains to solve the information-overload problem. Recently, trust-based recommender systems have incorporated the trustworthiness of users into CF techniques in order to improve recommendation quality. Some researchers have proposed rating-based trust models to derive trust values based on users’ past ratings of items, or based on explicitly specified relations (e.g. friends) or trust relationships; however, the rating-based trust model may not be effective in CF recommendations due to unreliable trust values derived from very few past rating records. In this work, we propose a hybrid personal trust model which adaptively combines the rating-based trust model and explicit trust metric to resolve the drawback caused by insufficient past rating records. Moreover, users with similar preferences usually form a group to share items (knowledge) with each other; thus, users’ preferences may be affected by group members. Accordingly, group trust can enhance personal trust to support recommendations from the group perspective. We then propose a recommendation method based on a hybrid model of personal and group trust to improve recommendation performance. The experimental results show that the proposed models can improve the prediction accuracy of other trust-based recommender systems.

Introduction

Recommender systems have been, and continue to be applied in various applications to support item (e.g. movies or music) recommendation [19], [30], [34], [46], [47], [52], and to solve the information-overload problem by suggesting items of possible interest to users. Even in a knowledge-intensive environment, recommender systems are able to support knowledge workers as they perform tasks, by recommending appropriate documents to suit their task needs. Of the various available recommendation methods, collaborative filtering (CF) [24] has been the most widely and successfully used method in various applications. It predicts user preferences for items by considering the opinions (in the form of preference ratings) of other similar (e.g. “like-minded”) users. Thus, personalized recommendations are made according to the preferences of similar users.

Recently, trust-based recommender systems [28], [31], [42] have incorporated the trustworthiness of users into CF techniques in order to improve the recommendation quality. These trust computation models [16], [21], [42] are used in trust-based recommender systems to derive trust values based on users’ past ratings of items. Such trust computation models can be classified into two categories: reputation trust and relationship trust [26].

Reputation trust is a quantitative assessment that allocates a trust score to a specific person by accumulating other users’ or a group of users’ trust scores on that person [9], [21], [42]. Some researchers call this global trust [9], [22], [42]. On the other hand, relationship trust is the trust between two users. One user trusts another based on past interactions or explicitly specified relationships [15], [16], [25], [28]. Some researchers call this personal trust, or local trust [13], [16], [40], whose value is limited between two users and diversified with different user pairs.

There are two categories of calculating trust scores (trustworthiness) between users. One category of trust-based system computes the trust scores based on users’ past ratings on items [42], while the other uses an explicitly specified trust metric to derive the trust values based on explicitly specified relations (e.g. friends) or trust relationships [14]. Users need to specify explicitly whom they trust and how much they trust each other.

O’Donovan and Smyth [42] suggest that if a user has usually delivered accurate predictions in the past, then s/he merits being called reliable and trustworthy. The accuracy of a prediction indicates whether the difference between a predicted rating given by a user (producer) and the real rating given by a target user is within a predefined error bound. A user is more trustworthy if s/he has contributed more accurate predictions than other users did; this trust model is reputation trust, and includes item level and profile level. The item-level/profile-level trust metric of a given user is derived by computing the ratio of accurate predictions that s/he has made to other users over a particular item/all items that user has rated in the past.

Massa et al. [37], [38], [39], [40] propose a relationship-trust recommender system based on a user’s Web of trust, which explicitly specifies the friends that s/he trusts. Their work, however, relies on the user’s explicit assignment of trust values, which are not easy to collect, and may create a heavy burden on users. In addition, Hwang and Chen [16] propose a relationship trust metric to derive the trust value between two users by calculating the ratio of accurate predictions over all co-rated items, i.e. those items that have been rated by both users. Their proposed relationship trust metric is personal trust, and is more personalized than the reputation trust metric.

The rating-based trust model derives trust values between users based on their co-rated items. If two users have very few co-rated items, the trust value derived from the ratings of their co-rated items may yield misleading trustworthiness between those users. The rating-based trust model, therefore, may not be effective in CF recommendations due to unreliable trust values derived from insufficient past rating records.

Although conventional trust-based CF systems have proposed rating-based trust models, or explicitly specified trust metrics to derive the trustworthiness of users, they do not investigate the combination of the rating-based trust model with an explicit trust metric. In this work, we propose a personal trust model that adaptively combines the rating-based trust model and explicit trust metric to resolve the drawback caused by insufficient past rating records. We derive the trust values between two users based on their explicitly specified role relations. Such explicit relationship trust can complement the traditional rating-based trust model in improving the reliability of trust values. The proposed model adaptively adjusts the relative importance of rating-based trust and the explicit relationship trust based on the number of co-rated documents between two users.

Moreover, users with similar preferences usually form a group to share items (knowledge) with each other; thus, users’ preferences may be affected by group members. Accordingly, group trust can enhance personal trust to support recommendations from the group perspective. Nevertheless, conventional trust-based CF systems do not address trust computation by considering both personal and group trust. Therefore, we propose a hybrid trust model which integrates personal and group trust in order to improve the performance of collaborative filtering. From the group-based trust metric we can obtain recommenders that are trustworthy from the group’s point of view. Such a group perspective may be important because it can complement the trustworthiness of the personal perspective, in particular, when an individual is not sure who to trust. In the group-based trust, we define a role-weight for each user to represent his/her degree of importance within the group. By adopting the role-weight value, the group-based trust can be aggregated from group members’ trust values. On the other hand, the group-based trust focuses on the majority of the group’s opinions, which might ignore the personal perspective. Accordingly, our proposed hybrid trust model combines personal trust and group-based trust models to integrate the merits of both perspectives. The trust values derived from our trust models are regarded as weightings in the collaborative filtering (CF) method to identify the trustworthy recommenders for predicting document ratings. Our experiment results show that the proposed trust model can improve the prediction accuracy of the CF method, compared with other trust-based recommender systems.

This paper is organized as follows: We present related work in Section 2. An overview of our trust computation models from the personal and group perspectives, and recommendations based on these trust models are presented in Section 3. The experiment results and evaluations are discussed in Section 4. Conclusions and suggested future work are presented in Section 5.