Community Driven Search Engine Based on Community's Proxy Server Log

抄録

In this paper, we are introducing a new method to improve search engine capabilities by user preference achieved with the help of community's proxy logs. The goal is focused to build a custom search engine that providing community-specific results. To achieve such search engine, we use proxy server logs from Network Operation Center of EEPIS-ITS and fetch the url and user field as a base for our work. Then, we use tf-idf algorithm to convert those textual data into a machine friendly numerical data. To find topics based on those url, we cluster it into 10 or more preferable clusters using k-means algorithm. Getting the result of that method, then we crawl the title and meta information from all of the clustered url to find the actual topic. Those result, finally would be our base to create the search engine. Lastly, we use vector space model to provide a search result from user's query.