Research On User Choice Behavior And Product Recommendation Of Fresh E-commerce Based On Online Reviews Mining

With the promotion of "Internet +" and the development of users' consumption habits,fresh food e-commerce platforms have become an important online marketing channel for fresh products.However,due to the perishability of fresh products,the difficulty of cold chain logistics,the inconspicuous brand,and the low degree of standardization,consumers still lack trust in the fresh products purchased online.Information before the purchase of fresh products in online consumption scenarios has higher value to users.Reviews,as a kind of user-generated content,are an important way to disclose business reputation as well as service/product information.It becomes an effective information acquisition source for users and affects users' purchasing decisions.To achieve precision marketing,fresh food e-commerce companies need to efficiently use user-generated information,timely and accurately mine user needs,and help users solve the problem of "information overload" and find target products accurately.This thesis is based on online review text mining technology,mathematical modeling,empirical research,and recommendation algorithms,to study how to accurately extract data,understand the concerns and needs of users for fresh products and services,explore the impact of implicit feedback information in comments on user choice behavior from a quantitative perspective,and implement intelligent fresh product management and product marketing recommendation models for fresh food e-commerce.This thesis provide theoretical and practical guidance for mining user needs and preferences,realizing precise marketing and personalized product recommendation.Firstly,considering the social impact utility of comments,the research on the hidden needs and preferences of online heterogeneous users was conducted.Using Internet big data to mine user demand characteristics in comments through text mining technology.Then combined with the attributes of the user and the attributes of fresh agricultural products(origin,weight,price,etc.),considering the impact of implicit feedback information on the user's purchasing decision behavior,a Mixed logit model that integrates the social impact of comments is constructed,and the analysis of online selection behavior of users is carried out.Finally,an example analysis is carried out through the actual fresh food e-commerce data.The results showed that: fresh product attribute variables(explicit demand)cannot fully capture users' needs,and more invisible needs are implicit in the reviews of other users,which affects users' choice behavior.Compared with the model that does not consider the impact of reviews,the user online choice model with considering the effect of comments has the better explanatory ability,which verifies that the proposed model has practical value for studying the choice behavior of online users.Secondly,the fresh product recommendation algorithm based on revised users' rating and Item-based collaborative filtering was studied.The characteristics of fresh products and their online transaction features were considered,a product feature extraction method based on improved part-of-speech path extraction algorithm and a feature-based sentiment analysis method based on sentiment dictionary were combined to fine-grained differentiate of ratings.Then the popularity time decay factor and seasonal sensitivity factor were used to improve the score prediction function,and a fresh product recommendation algorithm that combines revised ratings and Item-CF algorithm was established,thus generated recommendations.Experimental results showed that the proposed ratings revised method can effectively distinguish users' preferences and has a good effect on improving the accuracy of recommendation results.Compared with algorithms that do not consider rating revision,the proposed algorithm improved the diversity and stability of recommendations.Besides,through considering the popularity decay factor and the seasonal sensitivity factor of fresh products,the accuracy and diversity of recommendations were effectively improved.At last,compared with User-CF algorithms,the Item-CF algorithm is more suitable for the recommendation of fresh products.This study enriches the application fields of the traditional discrete choice model and collaborative filtering recommendation algorithm.At the same time,this study provides ideas for fresh food e-commerce platforms about the decision-making of precision marketing,business strategy,and personalized recommendation.