ARES：Autoregressive Emotion-Sensitive Model for Predicting Sales Performance

Along with the vigorous development of Web 2.0, lots of comments that represent the voices of customers appeared on the Internet, and the general public's sentiments toward products are increasingly influenced by the underlying viewpoints. Therefore mining the sentiment information from reviews would produce practical values for predicting sales performance and adjusting market strategy. Aiming at this problem, based on the result of the analysis on the characteristics of online book reviews, it proposes a sentiment analysis method. First, a polarity word dictionary is automatically constructed by the part of speech list and the prefix list. Afterwards the sentiments in the reviews can be extracted based on the polarity dictionary. Finally, the paper presents an ARES (autoregressive emotion-sensitive model), to utilize the emotion information acquired by the sentiment analysis method for predicting sales performance. Experiments are conducted on a book data set. By comparing the ARES with alternative models that do not take sentiment information into consideration, as well as a model with a different sentiment analysis method, the results, on the one hand, indicate that our sentiment analysis approach could generate a well summary of the review itself, and on the other hand, confirm the effectiveness of the proposed prediction model.