Information Retrieval Ranking Using Machine Learning Techniques

Information retrieval is the research area in which many researcher have been done and many are still going on. The rapidly growing web pages make it very crucial to search up to date documents. In continuation of research works on learning to rank, this research focuses on implication of machine learning techniques for IR ranking. SVM, PSO and hybrid of both are the main techniques implemented for IR ranking. In case of SVM, selecting appropriate parameters is difficult, but it gives potential solutions for the ranking. One of the optimization methods i.e. PSO is easy to implement and has global search capability. Thus to find the fitness function to optimize the ranking of document retrieval Hybrid SVM-PSO model is proposed. After the comparative study it has been calculated that the ranking parameters gives best result for RankSVM-PSO over RankPSO and RankSVM. The result has been calculated based on single term queries and multi-term queries. The study shows RankPSO gives the better result than RankSVM and RankSVM -PSO gives better result than RankPSO, so it has been concluded that RankSVM-PSO gives best result among the three techniques.