As the size of the real-time feature vectors are increasing day-by-day, finding an essential key feature sets for cross domain classification problem is difficult due to large data size and sparsity problems (missing values and imbalance). Traditional word embedding and feature selection models use limited sized data and dimensions for feature ranking and classification process. In this paper, a hybrid cross domain classification based feature selection is proposed in order to improve the efficiency of aspect sentiment classification on large databases. Experimental results show that the proposed cross domain feature selection based classification approach has better overall true positive than the conventional approaches.

As the size of the real-time feature vectors are increasing day-by-day