Analysis And Research Of Data Of Mobile Terminal Based On Data Mining

With the rapid development of mobile Internet, there are more and more products and technologies emerging in the world. As one of the most important Internet service in the traditional Internet age, search presents a rich diversity in the mobile Internet age. The search engine records every user’s usage data as many people uses it, including search time, query, the rank of clicking and the frequency of clicking, and we call all these records as search logs. The search behavior of users generates amount of data in the both traditional and mobile Internet age. And finding the valuable behaviors of users plays an important role in the development of search engine and enterprise.As one of the most popular technology, data mining is widely used in all walks of life. Generally speaking, data mining includes four kinds of task:cluster analysis, predictive modeling, association analysis and anomaly detection. There are lots of automated mining tools emerging because of the widely using and popularity of data mining, and weka which is mentioned in this paper is an example of open source data mining tools. It is very difficult to detect the potential models hiding in the search logs for the general statistical analysis tool, so the data mining is an effective way to detect these models.According to the above two backgrounds, we use both statistical analysis and data mining to research the search logs, and one thing should be mentioned that the data used in this paper is provided by Sogou company. After a preliminary analysis, we design a search logs analysis system based on data mining technology. We make an analysis on the fields of session, query and URL by statistical techniques, and mainly for the number of independent users in a certain period of time, the average searching times of users, the popular query words and the most clicked URL; in the aspect of data mining, we use a query proximity algorithm based on user clicking combine with hierarchical clustering to recommend the relational query words according to the query word input by users, and this will improve the user experience.