Potential Customer Identification Method Research Based On User-generated Content

In recent years,the social media platforms such as online forums,the micro-blog facilitate people to communicate with each other,and then create massive user-generated content which implies user's purchase intention to a certain extent and provides an alternative channel for enterprises to obtain their potential customers.However,the massive and unstructured user-generated content and the sparse distribution of potential customers causing great challenges for potential customer identification based on user generated content,which makes it difficult to analyze and identify potential customers from social media.There are few researches on potential customer identification based on user generated content,especially lack of research in the context of Chinese user generated content.Therefore,this thesis focuses on the characteristics of potential customer identification based on user-generated content,and takes the dataset from autohome.com as an example,the potential customer identification research based on user-generated content is conducted.First,this thesis analyzed the characteristics of the potential customer identification based on the user generated content,arranged the process of potential customer identification based on user generated content,constructed a framework for potential customer identification based on user generated content;Next,according to the text features commonly used in existing research and the characteristics of potential customer identification based on user generated content,the potential customer related features are extracted,a potential customer identification effective feature set is built;Then,aiming at the problem of high imbalance in experimental dataset,combined with Stacking ensemble learning classification algorithm,a Stacking classification algorithm specifically for imbalanced datasets is proposed to identify potential customers.Finally,the control experiments are set up to verify the actual effect of the Stacking classification algorithm specifically for imbalanced datasets proposed in this thesis,the experimental results show that this algorithm can be effectively used in potential customer identification.At the theoretical level,this thesis not only expands the research field based on user generated content,but also has a certain reference significance for the research on the same types or field,such as product demand perception.At practical application level,this thesis provides new ways for enterprises to acquire potential customers,reduces the cost of acquiring potential customers,and improves the efficiency of potential customers identification.