Emotion based social media text classification using optimized improved ID3 classifier

Emotion sensing is a complicated task of machine learning technology. That belongs to the Natural language processing (NLP) branch of artificial intelligence (AI). It is a broad domain of learning and analysis, the text classification and their emotional orientation discovery is also part of this domain. In this work the emotion based text classification is introduced. Thus previously developed data classification techniques are evaluated and a decision tree based text classification model is proposed. The key issues in traditional ID3 and its variants are to select optimal attribute from the training dataset for optimized learning. Also the performances of the previously developed algorithms are based on the size of training set which is another key issue. In this work the algorithm is designed such that its performance is almost independent of the size of the training data (size with minimum limit) as well as it tries to select more optimal attribute for decision tree. Therefore to improve learning ability modification on the attribute selection process is performed first. After improving the classifier's learning ability, the modified model is used with the NLP parsing tool for analyzing the text sentiments. Here the classification nature is binary classification. Additionally the comparative performance is also measured with the traditional ID3 algorithm and its variant classification algorithm. The experimental result based on parameters accuracy, error rate shows that the proposed technique outperforms as compared to its similar techniques implemented and gives more accurate sentiment orientations.