Comparative Analysis of Machine Learning Algorithms for Sentiment Analysis of Multilingual Nigerian Social Media Comments

Abstract

This study probes into sentiment analysis within the multilingual landscape of Nigerian social media comments by using machine learning algorithms as computational tools. Nigeria has various languages, creating a complicated scenario for understanding sentiment dynamics in digital discourse. This research developed sentiment classification models across languages such as English, Hausa, Yoruba, Nigerian-Pidgin, and Igbo. The choice of machine learning algorithms utilized in this paper was driven by algorithms suitability for the task, diversity of languages, and dataset characteristics. By utilizing machine learning algorithms such as Logistic Regression, Support Vector Machines (SVM), Random Forest Classification (RFC), and Long Short-Term Memory (LSTM), the study aims to provide insights into the varieties of expressions and sentiments in Nigeria’s social media. The research explores the domains of natural language processing and socio-linguistics. In summary, the research finds that SVM and Random Forest are the most functional for the combined multilingual dataset, giving a superior precision of 76% and recall of 72% each. For individual language datasets, LSTM, due to its advanced sequence modeling capabilities, outperforms the other machine learning algorithms with model accuracy scores of 58% to 90% and 72% for the combined language datasets.