CNN-BPSO Model for Multi Classification of Tweets
Manjubala Bisi, Ankam Divija, Sowmya Namala, Rohan Sarap · 2020
Twitter is a widely used platform for sharing views and opinions about every ongoing topic in public. A huge amount of opinion oriented data are posted on this platform so that every other person can know and respond to it. Classification of tweet data is a recent field of research. Researchers are developing different types of models for classifying tweet data into several classes. In this paper, Convolutional Neural Network (CNN) is proposed to classify tweet data into four different classes such as business, entertainment, health and politics. The performance of CNN depends on different hyper parameters such as input batch size, learning rate, number of filters, filter size and architecture. Binary Particle Swarm optimization is used in this work to optimize the hyper parameters of CNN. The proposed CNN-BPSO approach is evaluated over a real twitter data set and experimental results show that CNN with BPSO is able to provide better accuracy than CNN model.