Purpose

Determine if Ordinal Hyperplane (OHPL) combined with Long Short Term Memory (LSTM Recurrent Neural Network) out-performs the state-of-the-art alternatives in predicting multi-level ordinal classification.

Existing work reduces ordinal classes to binary classes by grouping class labels and apply Machine Learning methods like Support Vector Machine (SVM) and Naive Bayes classifier to achieve only binary prediction like pain/no pain, positive/negative, true/false.

Business Merit

Accurately predicting multi-class user rating based on text review can offer better insight for user preference/interest. Predictions can be used to draw better inferences and conclusions, as existing solutions only offering binary prediction results in a generalized sentiment prediction. Our proposed prediction model can be used to determine ratings from different text sources such as Facebook and Instagram on unrated reviews/sentiments. Predicted data is helpful for high ranking executives in their decision making process to improve product offerings and improve customer satisfaction.