Multiclass Text Classification and Analytics for Improving Customer Support Response through different Classifiers

Abstract

In any industry, Customer Relationship Management (CRM) is a very important aspect of the business. In a complex business environment, providing an efficient customer support service is always a challenge. Customer reports the issues/defects in the system to the vendor by sending emails or by creating a ticket in CRM tools like Salesforce.com. The content of such reports includes detailed technical problems or complex workflow issues due to system failures. In the industrial automation systems, a commissioning engineer or a field operating engineer generally reports such issues. Understanding and responding to the customer issues/defects and providing quick customer support is not an easy task. These CRM tools are not sufficiently astute to classify the defects into predefined classes. Text classification techniques are used to automatically identify and categorize the defects from the text messages. In this paper, five different machine learning classifiers (i.e. SVM, MNB, Decision tree, Random forest and K-nearest neighbors) are applied to perform multiclass text classification. The text messages are classified into predefined twelve technical system defects. The comparative analysis of five different classifiers on Customer Support dataset shows that the Support Vector Machine (SVM) has a better accuracy score in identifying the defects.