An intelligent system for relevant information retrieval and personalized recommendations

Abstract

In this thesis, an architectural framework for an Intelligent System which provides features for Relevant Information Retrieval and Personalized Recommendations has been proposed and implemented. This system provides facilities for effective access to user interface and search engine, web pages retrieval, fuzzy association rule generation, classification of web pages, Knowledge acquisition, Consultation with Domain Expert, Rule Management, Clustering, and User Profiles analysis for Relevant Information Extraction and Web Personalization. In this work, keywords are used to get the initial list of web documents. These web documents are then preprocessed using a newly proposed Anova-T Residue Statistical Classifier algorithm which is used for user profile analysis. This work also proposes a Fuzzy-D Discretization algorithm for feature selection in order to perform effective classification while selecting the relevant user features and to reduce the classification error. This proposed classification algorithm provides better classification accuracy when it is compared with the existing classification algorithms. Moreover, new algorithms called Ontology based collaborative filter algorithm and semantic based page ranking algorithm are proposed and implemented in this research work in order to retrieve relevant information with help of already prepared user profiles. Finally, this thesis proposes a new algorithm called Fuzzy- Temporal Association Rule Mining Algorithm (FTARM) to perform effective web personalization. Since a Fuzzy logic based temporal association rule mining approach is used in this work, the relevancy is increased when web pages are retrieved by users. Using the proposed FTARM algorithm and user profile analysis, personalized web pages are recommended to the users and at the same time these algorithms help the user to perform relevant information retrieval. newline newline newline