A Time-Based Online Customer Behaviors Analysis Model

Abstract

With the advent of the big data era, reviews are common when shopping online. To better understand the connection between reviews and the quality of the real product, we have to use mathematical language to simplify the problem. This relationship can help both the buyer to get what they want and the firms to make better products with attractive features. To help us understand how the indicators like star ratings can reflect the product's popularity, we construct a series of models to illustrate them. We first use the NLP method to process text type data, such as product review information, and convert these text data into sentiment scores, with the help of the k-means++ algorithm, we cluster the data into 3 categories. These categories can help the company to track the performance of their products by simply calculating which category the product belongs to. Next, a Time-based Reputation Judgment Model combined with AHP and Markov Chain Model was established to describe the change of reputation over time. In this way, a company can easily predict the future trend of their merchandise sales.