Hybridized Dimensionality Reduction Method for Machine Learning based Web Pages Classification

Feature space high dimensionality is a well-known problem in text classification and web mining domains, it caused mainly by the large number of vocabularies contained within documents. Several methods were applied to select most useful important features over years; however, performance such still improvable from different aspects as computational cost accuracy. This research presents an enhanced cosine similarity-based hybridization two efficient feature selection for higher performance. The reduced sets are generated using Random Projection (RP) Principal Component Analysis (PCA) methods, individually, then hybridized based on similarity values between features’ vectors. proposed method terms accuracy F-measure was tested dataset pages several term weighting schemes. As compared relevant results show significantly f-measure less set size. Index Terms— Cosine similarity, Dimensionality Reduction, selection, PCA, Projection.