关键词: 在线评价, 情感分析, 多属性购买决策, 后悔理论, 博弈交叉效率模型

Abstract: Purchasing is a complex behavioral decision-making process that often occurs in consumers’ daily lives. When consumers make product purchases, especially when they first purchase expensive, less frequent, or risky products, they will collect extensive information to make purchase decisions. The rich online review information of e-commerce websites becomes a potential resource for consumers to gather product information. Therefore, it is of great significance to explore the emotional attitudes in reviews and combine them with ratings to provide consumers with recommendations to assist in decision-making when shopping.
Existing studies can be divided into two categories, namely online review-based and online rating-based approaches to product purchase decisions, with the following four main problems. (1)The data type of indicators is single. While the reviews have a broader and more flexible affective expression effect, the ratings have a relatively bright emotional benchmark, which can better obtain the satisfaction of website users by combining the two kinds of data. However, most existing studies focus on only one type of data, ignoring the interdependent and complementary relationship between reviews and ratings. (2)Attribute weights are determined in an overly subjective way. On the one hand, the method of attribute weighting in the form of scoring by experts is highly subjective, and the reasonableness of the given weights is open to question. On the other hand, it is hard for consumers to assign attribute preference weights or expectation values in advance because it is difficult to quantify their preferences and expected needs into specific values in actual purchase decision scenarios. (3)The consumer’s regret psychology is ignored. Existing studies have focused all their attention on evaluating user satisfaction with products rather than consumer purchase psychology, and this satisfaction is hardly perceived by the consumers directly. Consumers can usually only perceive the idea of regret after purchase, especially if they miss a better choice. (4)Simplifying purchasing behavior into a selection process without considering the prices of the products. Purchase decision contains the most basic transaction process, which means that consumers need to pay the currency corresponding to the price to obtain the application value and usage experience of the products. Existing studies have considered only the choice problem but ignored the transactional attributes.
We present a novel product purchase decision method based on online reviews and ratings, considering consumer regret psychology and product prices. Firstly, we use a web crawler to obtain online evaluations of alternative products and process reviews and ratings separately. Secondly, after pre-processing the reviews, the sentiment dictionary of the product attribute domain is constructed using the improved character sentiment value calculation method. And the sentiment intensity of alternative product attribute reviews and ratings is calculated. Thirdly, we calculate the regret-rejoice values of alternative product attribute reviews and ratings, respectively, based on the regret theory, as the value of regret psychology will arise after consumers purchase the products. Finally, we take the regret-rejoice values of reviews and ratings as outputs and prices as inputs. The efficiency values are measured using a game cross-efficiency model with negative indicators to obtain the recommendation order of alternative products. In particular, when the efficiency values are the same, the lower price will be ranked higher.
The innovations are in four main aspects. (1)Considering both reviews and ratings in online evaluations to maximize the retention of valid information and obtain the user emotion. (2)Considering the regret psychology of consumers, and using the regret theory to measure the regret-rejoice psychology of consumers, which is more in line with the subjective psychological activities of consumers in the purchase decision process. (3)Considering the unique nature of product prices, the existing choice problem is refined into a product purchase decision process, which is more in line with the actual objective material activity. (4)The DEA model is used in the product purchase decision, as well as the combination of the base point method and the game cross-efficiency model to obtain the game cross-efficiency model that can handle negative indicators.
We demonstrate that the proposed method is clear, operable and has practical application value through the example of the purchase decision of 10 new energy vehicles of AutoHome, and provides a new idea for further research on the method of product purchase decision based on the online evaluations.
The shortcoming of this paper is that the given attributes of the platform are used directly without feature re-extraction. Future study can consider re-extracting the attribute features, so as to facilitate multi-platform data fusion. In addition, this paper and other literature assume that positive sentiment words in reviews have the same degree of influence on consumers as negative sentiment words when identifying sentiment intensity. However, negative sentiment words of the same sentiment intensity may have a higher impact on consumers, which is a direction for future research.

Key words: online evaluation, emotion analysis, multi-attribute purchase decision, regret theory, game cross-efficiency model