Research Of Trust Prediction Methods For C2C E-commerce

In order to combat speculative trade, reduce the shopping risk of the consumers, and enhance trust of the consumers in C2C e-commerce,trust prediction is significant to reduce the risk of a trade. Most of the existing trust models are based on the evaluations which are after a trade to calculate the trust value of a merchant. Those are the trust prediction methods which are after an event. However, it is more significant to predict the credibility of a merchant before a trade. This thesis presents a trust prediction method for a future trade and predicts the credibility of a merchant for the consumer before a trade. The main work and achievements in this thesis are as follows:(1) A prediction method which is for a future trade is proposed based on grey prediction theory. The method analyzes the trust value of a merchant in the past, regards the trust value as time series and uses the GM (1,1) model of grey prediction theory to predict the possible trust value of the merchant in the future. The experiment shows that the accuracy of GM (1,1) model is high for the sequence of trust value which is sustained growth. But, the accuracy of GM (1,1) model is lower for the sequence of trust value which is waved.(2) A method to calculate the trust value of a merchant is proposed based on multiple dimensions. Trust values in history are the premise of the trust prediction. This thesis analyzes the existing calculation methods, especially in C2C e-commerce, of trust value, and summarizes the features and disadvantages of them. Then, a method to calculate the trust value of a merchant is proposed which considers multiple dimensions such as feedback score, moment of feedback, amount of goods, credibility of evaluators, and penalty factor. In the view of credibility of the evaluators, a calculation method which is based on familiar and unfamiliar evaluators of the buyer is presented. In the view of the penalty factor, the amount of failures and the number of failures are considered to get the value of penalty factor by the weighted average. The results of the experiments show that the method of calculating merchant’s trust value is very effective to prevent merchants extracting trust value and punish the dishonest merchants. (3) An improved method based on the theory of Markov chain is presented. For the problem with lower precision of grey prediction method to analyze the data waved severely, a resolution is presented which revises the trust prediction value of the GM (1,1) based on the theory of Markov chain. The experiment shows that the improved method is effective for the data with the waved situation and the prediction accuracy is improved obviously.