GA based information retrieval system using dimensionality reduction techniques

In this paper, we propose a method of genetic algorithm (GA) for information retrieval (IR) based on Singular Value Decomposition and Principal Component Analysis. The main difficulty in GA based IR system is processing of high dimensional input strings, as affects the performance in terms of retrieval time. In proposed work, we tried to reduce the high dimensional input data to low dimensional in order to improve retrieval time. Singular Value Decomposition (SVD) and Principal Component Analysis (PCA) computations are done on input data streams to produce reduced rank matrix approximation and orthogonal transformations. With this reduced matrix we got promising results in terms of computation time when GA is applied for IR. Experiments were performed on sample dataset of 2500 input documents and weighted vectors are generated. Information retrieval is done using two techniques PCA-GA and SVD-GA with all classical matching functions. It is found that PCA-GA performs well as compared to SVD-GA in terms of computational time.