A dynamic online learning Framework using convolutionary Neural networks

Abstract

Neural Networks is useful in many applications including supervised and unsupervised learning. For data mining tasks, neural models are not usually used because they mostly give incomprehensible model and requires longer training time. In this work, We give a neural network learning algorithm which is able to give comprehensive model without utilizing excessive time for training. Convolution Neural Network is effective for performing text mining tasks which include feature extraction, feature classification an text prediction. It has been recently used for training classifiers which use character level textual data for automatically identifying features and high level concepts from the text. Pre-training and feature engineering is not required for training character level data. Text data is used for Online learning and predicting accurate results. In this work, we propose a prediction model that can greatly assist in predicting the relative and accurate text. The proposed approach is evaluated using real text data. We achieve 97%accuracy for the data set we considered and gives promising directions for future work.