Search Engines generally provide long lists of ranked pages, finding the desired information content from which is typical on the user end and therefore, Search Result Optimization techniques come into play. The proposed system based on learning from query logs predicts user information needs and reduces the seek time of the user within the search result list. To achieve this, the method first mines the logs using a similarity function to perform query clustering and then discovers the sequential order of clicked URLs in each cluster . Finally, search result list is optimized by re-ranking the pages. The proposed system proves to be efficient as the user desired relevant pages occupy their places earlier in the result list and thus reducing the search space. This thesis also presents a query recommendation scheme towards better information retrieval.