Design And Implementation Of Intelligent Analysis System Of Business Complaints For Mobile Operators

As mobile operators' business complaints become more and more complicated,relying solely on manual handling of complaints has problems of low efficiency and uneven processing quality.Complaint classification and solution recommendation are an effective solution technology.Complaint information is generally short text with sparse features.Traditional feature extraction methods cannot extract and describe the complete semantics of complaints,resulting in poor classification and recommendation effects.In response to the above problems,combined with the characteristics of two-way context in the BERT model when extracting features,a BERT-based complaint classification and solution recommendation method is proposed,and this method is applied to the realized business complaint intelligent analysis system.The main research work of this thesis is as follows:(1)Propose a complaint classification method based on BERT.First,use part of the complaint data to fine-tune the BERT model;extract the feature vector of the complaint based on BERT;finally,realize the classification by constructing a four-layer neural network model.Experiments show that the proposed method can greatly improve the precision,recall and F1 value of complaint classification.(2)Design a solution recommendation method based on the type of complaint and similarity.After categorizing the complaint category,recommend a custom solution based on the category;at the same time,based on the cosine similarity analysis between the BERT vector of the current complaint and the BERT vector of the historical complaints,recommend the Top N historical complaints with the highest similarity and their solutions to achieve the purpose of comprehensive recommendation and provide reference for handling complaints.(3)Design and implement an intelligent analysis system for business complaints.System development is based on Spring Boot and Bootstrap frameworks to realize five main functional modules: user information management,model management,feature extraction,classification and recommendation,and visualization of complaints hotspots.The classification and recommendation experiments carried out on the real complaint data set show that the proposed complaint classification method can effectively improve the accuracy of complaint classification,and the solution recommendation method based on the complaint category and similarity can also improve the application effect of the recommended complaint solution.The system can meet the needs of actual applications and improve the efficiency and quality of complaint handling.