Research On Properties Based Algorithm Of Product Ranking

With the rapid development of e-commerce, it has been a hot research topic that how to sort the associated goods retrieved from the keywords submitted by the users. Online shopping websites tend to provide rankings of goods from various dimensions for users to choose from the most suitable kind, such as comprehensive ranking, sales ranking, price ranking, good reputation ranking and so on.From the point of some commonly seen rankings of goods, this paper chooses the characteristics of sales, prices, number of comments, good reputation to study the rank algorithm.Through the analysis of the traditional ranking algorithm based on the rate of favorable comments, it is found that there exists some problems such as the accuracy of rankings when evaluating the data of a small sample with the traditional algorithm. This paper introduces the concept of Wilson confidence interval estimation, and proposes an improved algorithm which is based on the value of the lower bound of confidence interval to do the rankings rather than the rate of favorable comments. Besides, through the analysis the traditional algorithm based on the price, that using the characteristic of logistic to modify the price can solve the problems caused by the ranking of prices. What’s more, the effect of time on the results of rankings has also been taken into account when analyzing the algorithm based on the sales. Through the studies of the algorithm above, this paper aims at proposing a more reasonable algorithm which can do the rankings according to every single characteristic of the goods.Lastly, with the usage of rank aggregation, this paper provides a comprehensive ranking algorithm which takes all the factors into consideration. Upon testing and verifying the feasibility of the algorithm, we can do pre-treatment of the data collected by the web crawler, write a program to realize the algorithm and assess the model with NDCG. Experiments on real data set indicate that our improved method produces better ranking results.