Product sales forecasting using macroeconomic indicators and online reviews: a method combining prospect theory and sentiment analysis

Macroeconomic conditions and users’ word of mouth have significant impacts on the purchase decisions of consumers, and they can be potentially used to conduct better sales forecasts, but study on this aspect is relatively scarce. In this paper, a novel method for forecasting product sales based on macroeconomic indicators and online reviews is developed. Firstly, an algorithm is given to select proper macroeconomic indicators to capture the long-term trends of sales. Subsequently, an algorithm for sentiment analysis is given to convert textual online reviews into numerical digits, and the word-of-mouth effect is calculated by incorporating the data related to online reviews (e.g., ratings, browsing numbers, and approval numbers). The sentiment index of word-of-mouth effect is measured based on the prospect theory, which can accurately reflect the phenomenon whereby negative reviews seriously affect the purchasing decisions of consumers. Further, according to the selected macroeconomic indicators and the obtained sentiment index, a logarithmic autoregressive model for product sales forecasting is constructed, and the model parameters are estimated by the Adam optimizer. Finally, experimental studies on forecasting the sales volume of the Audi A6L in the next three quarters are conducted. The experimental results show that the performance of the proposed method is significantly better than the existing methods.