Multimodal sentiment analysis using reliefF feature selection and random forest classifier

Multimodal sentiment analysis becomes an emerging research topic due to the rapid growth of modality and diversity of social data. The primary objective of this research paper is to develop an effective feature selection algorithm to determine the optimal features for further improving the performance of multimodal sentiment analysis. Initially, the data were collected from YouTube dataset. Then, feature extraction was carried out on the collected data utilizing Mel-Frequency Cepstral Coefficients (MFCCs), linear predictive coefficient, spectral centroid, spectral flux, Local Binary Pattern (LBP), Histogram of Oriented Gradient (HOG), Latent Semantic Analysis (LSA), and Term Frequency-Inverse Document Frequency (TF-IDF) to extract the feature values from the collected data. After feature extraction, reliefF feature selection algorithm was used for choosing the optimal features. At last, random forest classifier was used for classifying the sentiments of speakers such as neutral, positive, and negative class. The quantitative analysis showed that the proposed system enhanced the classification accuracy up to 5.41% compared to the existing systems.