Performance study to explore machine learning based text classification method

Text Classification has become an essential feature in many classification applications such as text analysis, sentiment analysis and many more. With the advent of text based data obtained through Social Media, news articles, etc, there is a need for efficient machine learning methods capable of performing accurate textual classification in imbalanced data. The object of classification in this study has taken from consumer complaints data on financial products. The classification task is to help the financial organizations in identification, followed by providing customized solutions.

An extensive experimental text analysis is performed based on accuracy measure. The analysis is conducted using different text data representation, machine learning based classification tools with some balancing strategies aiming to find an efficient text classification models. This study involves experimenting the data with different machine learning models including Logistic Regression, Naive Bayes, Linear Support Vector Machine and the Random Forest and explored the best text classification model based on results.