Improved Density-Based Learning to Cluster for User Web Log in Data Mining

Abstract

The improvements in tuning the website and improving the visitors’ retention are done by deploying the efficient weblog mining and navigational pattern prediction model. This crucial application initially performs data clearing and initialization procedures until the hidden knowledge is extracted as output. To obtain good results, the quality of the input data has to be promisingly good, and hence, more focus should be given to pre-processing and data cleaning operations. Other than this, the major challenge faced is the poor scalability during navigational pattern prediction. In this paper, the scalability of weblog mining is improved by using suitable pre-processing and data cleaning operations. This method uses a tree-based clustering algorithm to mine the relevant items from the datasets and to predict the navigational behavior of the users. The algorithm focus will be mainly on density-based learning to cluster and predict future requests. The proposed method is evaluated over BUS log data, where the data is of greater significance since it contains the log data of all the students in the university. The conducted experiments prove the effectiveness and applicability of weblog mining by using the proposed algorithm.