MOHCOA: Multi-objective hermit crab optimization algorithm for feature selection in sentiment analysis of Covid-19 Twitter datasets

Abstract

The COVID-19 pandemic has led to a flood of data on Twitter, making it crucial to analyze public opinion. However, the large amount of data is challenging to manage. This paper presents the multi-objective hermit crab optimization algorithm (MOHCOA) to tackle this problem by improving the accuracy of sentiment analysis, selecting the best features, and reducing computing time. Inspired by how hermit crabs choose their shells, MOHCOA balances exploring new features and using known ones, which helps in better sentiment classification while cutting down on unnecessary data and processing time. Compared to other methods, MOHCOA is more efficient in selecting features and improving model accuracy. For the bag of words (BoW) set, MOHCOA narrowed features down to 2005, and for the BoW + COVID-19 keywords set, it chose 2278 features. When used with a random forest model, MOHCOA achieved a precision of 0.84, recall of 0.69, F1-score of 0.75, and accuracy of 0.83. This shows that MOHCOA is effective in managing large data sets, making it a useful tool for analyzing text and public sentiment during events like the COVID-19 pandemic.