Research And Implementation Of Intelligent Classification System For Telecom Customer Complaints

Customer complaints are centralized feedback on the multi-faceted problems of enterprises and potential needs,and their handling is directly related to the sustainable development of enterprises.With the rapid growth of the number and type of customer complaints,the intelligent complaint classification system has received widespread attention.However,there are still some problems with the existing intelligent classification system for telecom customer complaints:(1)There is a lack of high-performance customer complaint classification algorithm.First,the existing model pays little attention to data processing issues.Second,the existing model fails to make full use of the label constraint relationship of hierarchical multi-label text classification.(2)The existing system is not intelligent enough,and it is difficult for users to update and maintain data and models,and it is also difficult to adapt to complex and changeable application scenarios.In view of the above problems,the main research contents and achievements of this paper include.(1)This paper proposes a highperformance intelligent classification method for telecommunications customer complaints.In this method,a set of data analysis and processing plans is formulated based on the characteristics of telecom customer complaint data.Secondly,a hierarchy-associate multi-label text classification model is designed,which proposes gated-guided label coding of complex structure,associated learning of hierarchical information,and adaptive updating of multi-task loss weight with consideration of upper and lower classifiers,so as to make full use of hierarchical label constraint information.Finally,the model is combined with the traditional rule matching method to realize the multi-level decision-making intelligent classification method for telecom customer complaints,and the experimental results show that the performance of telecom customer complaints classification algorithm is significantly improved.(2)Based on the classification method of customer complaints,a highly intelligent telecom customer complaint classification system is constructed.The system integrates customer complaint data management,model management and classification service,especially provides rich data version management capabilities,flexible and controllable custom coding capabilities,and adaptive and manual model update capabilities.It covers all the process of model service update and landing,realizes the goal of assisting users to quickly update and maintain models,and improves the intelligence level of the system.