Efficient information retrieval using multitype feature coselection for clustering in heterogeneous database

Abstract

Information is quotknowledge communicated or received concerning a particular fact or circumstancequot. It is a sequence of symbols that can be interpreted as a message and helpful because it allows us to answer the who , what , where , when , and how many questions. Retrieving information from an unstructured source is a challenging job. newline Information Retrieval is a science of retrieving the relevant information from the unstructured collection of database. Feature selection methods and Clustering techniques improves the retrieval efficient. newlineFeature selection is one of the important and frequently used pre-processing techniques in Data mining. It reduces irrelevant, redundant and noisy data and brings the immediate effect for application, by improving the mining performance in accuracy and result comprehensibility. newlineA feature selection algorithm designed with different evaluation criteria, broadly falls into three categories: Filter model, Wrapper model and Hybrid models. It has been widely applied in text categorization and clustering. newlineFeature Selection has proven to be a valuable technique in supervised learning; compared to unsupervised selection. Supervised feature selection is successful in text categorizing of filtering the noise in most cases. newlineDue to the absence of class labels, clustering can hardly exploit supervised selection. Also to improve the clustering performance in the supervised feature selection, the intermediate clustering result is generated iteratively. In addition, a technique - Coselection is implemented in combining the similarities based on several types of features. newline newline newline