Application Of User Behavior Analysis In Search Engine

With the growing number of the Internet web information, web users began to have high expectations of the search engines function, intelligence and search efficiency, expecting too much accurate and reliable search results that are consistent with their individualized needs. Therefore, it is quite an important and valuable problem to strengthen and improve user tracking function of the search engine, identify user's behavior and demand characteristics from the information search process, and then to record, study and find the potential demand and preferences for users, understanding changes of correlative information user need timely.Based on the work principle of open-source Nutch, the paper realizes and designs a search engine, which has the function of analyzing user behavior. The basic idea is analyze the search engine logs to get users' preferences, and according to these preference judgments to enable the search engine filters and reorganizes results. It means that the search engine gets the user's preference from user's clicking behavior on the list of search results, and then adjusts and updates the original retrieval function with machine learning methods, in order that the interesting websites are placed in the front toward in the search results so as to enhance the search accuracy.In recent years, support vector machine algorithm is widely used on the issue of learning of retrieval function (ranking function) and performs excellently. In this paper, the algorithm is used for each word retrieval function that reflects the correlativity between words and documents. Then the inquiry documents are ordered from low to high score in accordance with the retrieval function. The test results shows that the algorithm basically achieved the expected goal and improved search function of search engines.The system is developing by Java and running on local area network and has the function of recording users' behaviors, analyzing users' preferences to learn retrieval function and updating automatically search results. The engine can keep up with user and discover alternatives of user's preference and update search results in time. Also, it improves intelligence and retrieval quality of general search engine. It has achieved the aim of expected study and practice and laid a basis for the further study web mining.