Tamil Nadu 2026 Assembly Election Prediction Using Machine Learning and Dravidian Social Media Sentiment Analysis

Abstract: Virtual election forecasting using machine learning represents a transformative approach in the domain of political data science. Predicting election outcomes with accuracy has always been a challenging problem due to the complex interplay of historical voting patterns, incumbency effects, party momentum, and evolving public sentiment. This paper presents a machine learning-based framework for predicting the Tamil Nadu Legislative Assembly Election 2026 across all 234 constituencies. The system integrates Tamil Nadu Assembly Election data spanning 1971 to 2021 with real social media sentiment derived from the DravidianCodeMix dataset—comprising 43,632 Tamil code-mixed posts. A Random Forest classifier trained on seven engineered features achieved 88.84% accuracy on the 2021 validation set, outperforming the Gradient Boosting baseline of 87.12%. Sentiment analysis using a LaBSE+SVM model (NAACL DravidaLangTech 2025) was applied to 2,560 party-matched tweets to produce a sentiment score per party. The final 2026 prediction assigns 133 seats to Dravida Munnetra Kazhagam (DMK)—crossing the 118-seat majority threshold—65 to All India Anna Dravida Munnetra Kazhagam (AIADMK), and 18 to Indian National Congress (INC), consistent with the 2021 electoral outcome trend.

Keywords: Election Prediction, Machine Learning, Random Forest, Sentiment Analysis, DravidianCodeMix, Tamil Nadu 2026, LaBSE, Gradient Boosting.