| In recent years,in the context of data as a key factor of production,the digital economy is the central driver of China’s high-quality economic development.E-commerce platforms,such as JD,Tmall,and Pinduoduo,have also witnessed rapid development as an important part of the digital economy.The development of e-commerce comes with the rise of shopping methods.The massive data accumulated by e-commerce platforms have become the major source of information for consumers to perceive online shopping risks.How consumers make choices when faced with online shopping risks that are different from traditional shopping is worth studying.Collected a total of 820smartphone product details and store information on Tmall and JD platforms,and more than 420,000smartphone online review data,helpful online reviews extraction methods,and online shopping risk assessment methods are proposed,a causal relationship model for consumer online shopping behavior selection is established based on the results of online shopping risk assessment by integrating online reviews,product parameters,and store information.The main research contents are as follows:(1)Research on the extraction method of helpful online reviews based on Random Forest.The influencing factors of the helpfulness of online reviews are divide into 11 information factors and 4normative factors according to dual-process theory;The online reviews are integrated into the structured data of easily accessed according to feature analysis and sentiment analysis methods.Constructing a random forest prediction model for the helpfulness of online reviews based on the JD reviews data.The R~2 is 0.855.This prediction model can be used to extract helpful reviews of Tmall,it is found that 24%of Tmall smartphone reviews are helpless reviews.(2)Research on online shopping Risk assessment method by integrating online reviews,commodity parameters,and store information.Firstly,the formation mechanism of online shopping risk is analyzed,the online shopping risk is divided into three dimensions:commodity risk,merchant risk,logistics risk,and an indicator system of online shopping risk is established.The online shopping risk is divided into four levels by using K-means clustering:low risk(I),medium and low risk(II),medium and high risk(III),and high risk(IV).The data is divided into a training set and test set at a ratio of 4:1,and the accuracy of online shopping risk assessment is verified by using Naive Bayes,Decision Tree,Random Forest,and Support Vector Machine classifier.The results show that helpful online reviews can improve the accuracy of risk assessment,and the classification accuracy of Random Forest is higher than the other three classifiers,and the prediction accuracy rate is 99%.(3)Research on consumer behavior selection based on Bayesian network.Considering the structural correlation and the influence between the indicators,the decision tree is used to deduce the division interval of risk indicators,the Bayesian network model of"Consumer Behavior-A three-dimensional online shopping risk-risk indicators"is selected to develop a dynamic model for consumer behavior.Sensitivity analysis shows that the influence of commodity risk,logistics risk,and merchant risk on consumer behavior selection decreases in turn,the impact of logistics service score and the emotional score of logistics service review on consumer behavior selection is greater than merchant risk;Causal reasoning and diagnostic reasoning show that high risk is a sufficient condition for consumers to choose to not buy products.In this paper,the helpful online reviews extraction method proposed can effectively alleviate the problem of information overload;Online shopping risk assessment methods contribute to preventing and resolving online shopping risks,purifying the online shopping environment,and improving consumer satisfaction;The law of consumer behavior selection can provide a practical way for e-commerce platforms to explore consumer behavior characteristics and promote consumption upgrades. |
PDF Full Text Request