Elections are a critical aspect of democracy, where voter sentiment and political party support significantly influence outcomes. This study aims to predict election results in Tegal City using machine learning models, specifically Neural Networks, Random Forest, and Naive Bayes. Each algorithm was applied to a dataset containing demographic, polling, and Sentiment data to analyze political party support. The research revealed that Neural Networks outperformed other models in terms of accuracy (92%) and F1 scores for both positive (91%) and negative sentiments (92%). Random Forest and Naive Bayes, while effective, displayed lower overall performance. The findings highlight the value of utilizing advanced algorithms for local election sentiment analysis to help candidates adjust campaign strategies. This approach enhances understanding of voter behavior and supports more informed decision-making processes for the public and policymakers