Feature Selection for Ranking using Heuristics based Learning to Rank using Machine Learning

Abstract

Machine Learning based ranking is done every filed. Ranking is also solved by using (LTR i. e. learning to Rank)techniques. In this work, we propose a Heuristics LTR based models for information retrieval. Different newalgorithms are tackling the problem feature selection in ranking. In this proposed model try to makes use of thesimulated annealing and Principal Component analysis for document retrieval using learning to rank. A use ofsimulated annealing heuristics method used for the feature Selection to test the results improvement. The featureextraction technique helps to find the minimal subsets of features for better results. The core idea of the proposedframework is to make use of k-fold cross validation of training queries in the SA as well as the training queriesin the any feature selection method to extract features and only using training quires make use of validationand test quires to create a learning model with LTR. The standard evaluation measures are used to verify thesignificant improvement in the proposed model. Performance of proposed model are measured based on predictionon some selected benchmark datasets, Improvement in the results are compare on recent high performed pairwisealgorithms.