A Systematic Review of Predicting Elections Based on Social Media Data

• Python
• Machine Learning

In recent years, social media data has been used for election monitoring and prediction, especially Twitter data. Twitter has emerged as a popular communication channel between leaders of contesting parties and voters. During election campaigns and elections, both contesting parties and voters express their opinions on social media websites generating huge amount of unstructured data. This data is valuable for contesting parties and voters, one important use of this data being election predictions. Some early work in election prediction includes vote share prediction using mere volume of tweets showed that only volume of tweets with mentions of political parties can represent successfully election polls. They suggest Twitter as an acceptable real time indicator of public sentiment towards political parties. One of their limitations was that they were only able to measure overall voter?s sentiment rather than sentiment towards specific political topic. In this project we aim to eliminate this gap and classify each tweet on the basis of context or topic. Main contribution of this project is based on sentiment approach is that, it captures word relations and co occurrences, specially using rich corpus in short length tweets, and hence provides a better estimate of sentiment polarity and score of tweets for election prediction.

Project Sample Pages :