Hybrid optimization enabled Random multimodal deep learning for sentiment rating prediction

Abstract

Sentiment analysis is the most basic and imperative work in mining the preference of user interest. In this work, a deep model with optimization, named “Chimp Whale Optimization Algorithm-based Random Multimodal Deep Learning” is devised for sentiment rating prediction. The process of tokenization, which divides the entire document into small units using Bidirectional Encoder Representations from Transformers (BERT) for better processing, is where the input review data is initially given. Aspects from review data and aspect term extraction are completed for mining. Additionally, Random Multimodal Deep Learning is used to forecast the sentiment rating. The ChWOA is used in this case to combine the Chimp Optimization Algorithm (ChOA) and the Whale Optimization Algorithm (WOA). With a precision of 93.1%, recall of 94.4%, and F-measure of 93.8%, the ChWOA-based RMDL demonstrated better efficiency.