Predicting Sales Performance Based on Polarity Sentiments of Online Reviews and Manifold Dynamics Method

Online reviews provide consumers with information about products and services that may affect their purchasing decisions. As such, the customer attitude in the reviews play a important role to product sales. In this study, new sentiment prediction method is presented to enhance the forecasting accuracy by utilizing review dual-sentiments and employing non- linear manifold dynamics algorithm. Not only this method can extract and separate the polarity sentiment factors from the content of each online review, but also can remove the seasonality of products and mix the different types’ data. Extensive experimental results validate the effectiveness of our proposed method.