Design And Implementation Of User Reputation Evaluation Algorithm Based On Median

With the development of social network, people think more and more highly of knowledge sharing, web-based rating systems become popular in the field of e-commerce and business review website. In these websites, two kinds of entities exist ubiquitously:user and object. Users can rate for the objects, then system will calculate the final scores for the corresponding objects based on the user ratings, and the scores will decide objects’ranking. Obviously when user chooses a kind of product or service, his/her decision will be influenced by the rating information, therefore, these systems are vulnerable to the spammers. During recent years, researchers tried their best to improve the accuracy of recommendation systems. It’s a hot topic in the flied of data mining.An outstanding algorithm should be robust enough to withstand the attack from the spammers, and convergent. For this purpose, nowadays many researches induce the concept of user reputation. Algorithms will adjust user’s rating weight by measuring his/her reputation, thereby reduce the influence of spammers to object’s final ranking.This paper will look into the algorithms in this field, then come up with a more robust and efficient algorithm, thus increase the accuracy of rating system.The main works of this paper are listed as follows:●Giving a brief survey of the current research about recommendation systems, mainly elaborate two types:content-driven and user-driven.●Introducing the existing user-driven algorithms, analyzing the advantages and disadvantages of various algorithms.●On the base of experiments and analysis, using median’s character of unsusceptible to extreme ratings, propose a median based user reputation algorithm, then applies the algorithm on various real data sets and compares the output with existing algorithms.●Applying the preprocessed data to SVD++. Comparing with the original data set’s RMSE value, our algorithm achieves a better result, which shows the effectiveness of the proposed algorithm.