Text Classification using Convolutional Neural Network

Abstract

Rapidly increasing in the digital data led to large volume of data. Today’s world almost 80 percent of data is semi-structured or structured data. There is big issue to analyzed textual data from huge amount of data. Text mining is used for extracting interesting patterns from large number of textual documents. Text mining techniques included clustering, text pre-processing and classification. This paper focused on text classification technique of text mining. Applications namely search engines, newspapers and e-commerce portals classify their content or products for easy searching and navigation. This paper presents convolutional neural network (CNN) with Word2Vec word embedding technique for text classification. The proposed approach is tested on seven benchmark datasets with varying classes from 2 to 14. Experiments are conducted to identify suitable parameters of CNN such as batch size, epochs, activation function, dropout rates and feature maps values. Accuracy obtained by CNN model is better than other machine learning techniques such as support vector machine, naïve bayes for all datasets. Accuracy obtained by CNN model is closest from accuracy obtained by literature work for all datasets.