User Satisfaction Evaluation Of Search Engine Based On Click Log

How to evaluate search engines is an important issue of search engine research. With the increasing of data scale, traditional search engine evaluation methods can not meet the requirements of current search engine evaluation. Since search engine user click log implicits users’judgment of search engine results, which is easy to access, using user click log to evaluate the quality of the search engines becomes a possibility.In this study, we research on automatic evaluation of search engines to improve search engine evaluation efficiency and shorten the search engine evaluation cycle, which can promote the improvements of search engine performance.This paper presents a search engine evaluation method based on user click log. It analyzes the interaction between users and the search engine, and extracts the characteristics which can reflect of the quality of the search engine, and then use these characteristics to evaluate the quality of the search engine. Based on the analysis of search engine user behavior and user click log, we select search result ranking, click rate and average browsing time as indicators to estimate user satisfaction.Then we use linear regression analysis, logistic regression analysis and BP neural network method to build search engine evaluation models respectively.Experiments show that BP neural network model is superior to the linear regression model and logistic regression model by comparing the mean square error and accuracy of the three models.