User Satisfaction Prediction Based on Adaptive Feature Interaction Extraction

The timeliness of user experience research for current telecommunication operators is limited, and they overly rely on expert experience in analyzing user satisfaction, lacking exploration of deep potential characteristics of users, thereby limiting the precise perception of user experience. Therefore, this paper is based on the user satisfaction prediction problem of a certain operator enterprise in Guangdong Province, and proposes a satisfaction prediction model based on adaptive feature interaction extraction (AFIE). This model obtains high-order interaction interactions between user features through multi-head attention gating mechanism and exponential neural network structure, improving the accuracy of user satisfaction prediction. After experimental verification on the operator enterprise user dataset and two publicly available telecommunications operation datasets in the province, compared with the baseline model, this method significantly improves prediction performance.