2016 Presidential Election Prediction using Twitter
Weifeng Li · University of Bridgeport ScholarWorks (University of Bridgeport) · 2016
Nowadays, data of social media websites are getting more and more popular to be used as one of the most important data source for the data mining from which we can find the useful and interesting patterns. In this project, base on twitter data set that I collect using twitter API, I performed the sentimental mining and topic modeling. In the data collection phase, I used keywords such as the candidates’ name to filter the related data decreasing the noise to the most extend. To accomplish the sentimental mining, I chose Naïve Bayes algorithm and Support vector machine Model(SVM) two of the most commonly used algorithms that can be used as the classifier in the sentimental analysis. Then I trained these classifiers using a data set which was also from twitter and was related to 2016 presidential election from Kaggle and made the predication using twitter data set that I collected. Besides, Latent Dirichlet Allocation model was used to fulfill the topic modeling analysis finding the most frequent topics from the data of presidential election related tweets. At last, I evaluated the performance of each classification algorithm.