Cross Project Validation for Refined Clusters Using Machine Learning Techniques

摘要

Clustering is used for discovering groups and identifying interesting distributions and patterns in the underlying data whereas classification is a technique used to predict membership for data instances within a cluster. Correct classification of similar users in a cluster helps in better prediction of web pages. In the past lot of work has been done on original web log data whereas in this paper we intend to apply classification on refined clusters by implementing Modified Knockout Refinement Algorithm(MKRA). This approach leads to the improvement in cluster quality and prediction accuracy. After refining the clusters using MKRA we apply different learning techniques on refined clusters. Various performance measures of learning techniques are evaluated and compared. These days the machine learning community is trying to get better solutions for improving classification accuracy by applying ensembled classification. We further intend to apply ensembling on the classifiers used in our model to observe the betterment in the classification accuracy performance.