ETPPDMRL: A novel approach for prescriptive analytics of customer reviews via enhanced text parsing and reinforcement learning

Abstract

Necessity is the mother of inventions likewise understanding the requirement of the customer is the key to create a successful product. Customer review is one of the vital essences of fine tuning the products towards the flawless ones. Extracting useful information from text reviews is the primary objective for prescriptive analyzing. Modern science introduced a number of novel methods such as Artificial Intelligence (AI) through Machine Learning (ML) and Data Mining (DM). The introduction of Artificial Neural Networks (ANN) is driving the predictive analyzing process towards more beneficial for both customers and the manufacturers. An exclusive text parser is introduced in this work to make more compatible inputs for Reinforcement Learning (RL). C4.5. Prescriptive Decision Maker is introduced to achieve higher Accuracy and Precision. The new modules ‘Exclusive Text Parser for Reinforcement Learning’ and ‘C4.5. Prescriptive Decision Maker’ are the functional blocks used to construct the proposed method named as Enhanced Text Parser for Prescriptive Decision-Making using Reinforcement Learning (ETPPDMRL). Exclusive Text Parser and RL based C4.5. Classifier are submitted here as the novel contribution. Amazon customer feedback dataset is used to evaluate the performance of the proposed method during the experiments. Benchmark metrics such as Accuracy, Precision, Sensitivity, Specificity, F-Score and Average Process Time are used to evaluate the performance of the proposed method.