Twitter sentiment analysis using adaptive neuro-fuzzy inference system with genetic algorithm

In recent decades, it is very challenging for the researchers to identify the users’ sentiments from the twitter data, due to unstructured nature, misspells, abbreviations, limited size, and slangs. In order to address these problems, a new system is developed in this research study for improving the performance of twitter sentiment analysis. In this research, the proposed system comprises of three major phases; data collection, pre-processing and classification of sentiments from pre-processed tweets. Initially, the twitter sentiment analysis was carried-out by using twitter-sanders-apple 2 dataset. Generally, the raw collected tweets contain more noises by means of positive emoji’s, URLs, stop words, negative emoji’s, which were essentially eliminated to achieve better classification. Finally, a hybrid classifier (Adaptive Neuro-Fuzzy Inference System (ANFIS)-Genetic Algorithm (GA)) was used to classify the twitter sentiment classes as either positive class or negative class. The ANFIS classifier was the fuzzy based ontology that was designed by actualizing and the GA optimizes the fuzzy principles in the ANFIS classifier. The experimental consequence showed that the proposed methodology enhanced the accuracy in sentiment analysis up to 5.5-6% related to the existing methodology.