IOSR Journal of Computer Engineering (IOSR-JCE) e-ISSN: 2278-0661, p- ISSN: 2278-8727Volume 16, Issue 4, Ver. I (Jul-Aug. 2014), PP 06-10 www.iosrjournals.org www.iosrjournals.org 6 | Page Neighbor-based Data Weight Collaborative Filtering 1 Xuesong Yin, 2 Jie Yu, 1 Rongrong Jiang *1 School of Information & Engineering, Zhejiang Radio & Television University, Hangzhou, China 2 Zhejiang Agriculture & Forestry University, Hangzhou, China 1 School of Information & Engineering, Zhejiang Radio & Television University, Hangzhou, China Abstract: Existing Collaborative Filtering (CF) based recommendation approaches suffer from the following issues: (1) the number of resources accessed and evaluated by each user is only such a very small part that leads to sparse rating matrix; (2) dynamic change of user interest makes recommended resources largely deviate from the need of the user. To address these problems, we develop a novel algorithm titled as neighbor- based data weight CF recommendation of learning resources (NARR). Firstly, the neighbor of the user or the neighbor of the resource is selected in terms of the rating matrix; secondly, we compute data weight for representing dynamic change of user interest; finally, we use neighbor relationship and data weight in the objective of CF-based algorithm to choose learning resources. Experiments results show the feasibility and effectiveness of the proposed method. Keywords: Collaborative Filtering, Data Weight, Similarity, Data Sparseness I. Introduction Development of information technology makes knowledge transfer from the traditional paper-based forms of communication transition to digital transmission. Internet's rapid development and growing popularity makes online information become an important source of knowledge. Distance education regarded as a network technology based education form can maximize the use of network teaching resources. Under the education mode learners do not have to be time and space constraints and carry out learning at any time and any place according to their own needs. However, the main problem faced by distance education is learning resource overload. With explosive growth in the number of resources, on the one hand, learners cannot find the resources they need to learn; on the other hand, they are often unable to distinguish the pros and cons of resources, which results in low learning resource utilization. Recommender system (RS) is the very promising approach to solve the problem of current information overload. Different with information retrieval, RS provides personalized recommendations based on user profile and preference. Accurate recommendations enable users to quickly locate desirable items without being overwhelmed by irrelevant information so that users are gradually dependent on the system. Thus, RS can not only improve personalized service for users, but set up long-term stable relationship with users and improve customer loyalty. RS can be considered as social networking tools that provide dynamic and collaborative communication, interaction and knowledge [1][2]. RS is inspired by human social behavior and covers a wide variety of applications. It is common to take into account the tastes, opinions and experiences of our acquaintances when making all kinds of decisions (choosing films to watch, selecting schools for our children, choosing products to buy, etc.) [3][4][5]. Decisions of users are modified according to their interpretation of the similarity that exists between user group of acquaintances. Specifically, a RS aims at improve a user with those items that might be of his or her interest. RS has three elements: recommended candidates, users and recommended methods. The general recommendation system model is shown in Figure 1. RS actively collects users’ information or users voluntarily provide their preferences to RS. Then, RS can use different recommended strategies to show items. For example, RS calculates personalized information and user data collected to obtain recommended results or directly bases on the knowledge database modeled to recommend those items. Users Preference collector Recommender provide preferences Preference database preserve search preferences Item database provide recommendation Figure 1. General model of recommender system