Research On Classification And New Class Recognition Of Complaint Text In Business

Currently,in order to adapt to market competition and meet the needs of customers,enterprises focus on the customer service and constantly improve their services to improve customer satisfaction.The customer complaints is an important way to find enterprises themselves with problems,also is an important direction to improve their services and customer satisfaction.However,when it comes to customer complaints problem,the enterprises mainly rely on artificial identification and classification,which have not only low-efficiency,high cost,and low consistent results but also hard to identify the new category appeared in the process of classification.If customer complaints can not be handled timely and accurate,it may influence the subsequent processing and decision-making of the relevant departments,and even cause immeasurable loss to the enterprise.Therefore,the use of intelligent classification of customer complaints and identification of new categories have become urgent problems.This paper aims to establish a business complaints classification model and to identify the new category appeared in the process of classification.Firstly,it introduces the related concept of text classification and text classification technology of home and abroad as well as summarizes the advantages and disadvantages of the current classification methods from the background of complaints classification.By summarizing and analysising the complaints of business content and putting forward the two-dimensional classification system.the research divided the text classification system into business dimension and life cycle dimension.Besides,classification and new type recognition model have been constructed based on business dimensions,as well as the classification model been constructed based on life cycle dimension.Secondly,given that the Chi-square Statistic methods tend to choose negative words,so the research works on the improvement of Chi-square Statistic method,which means reducing the weight of negative words in the feature selection algorithm and minimizing the impact on the model.Then,on the base of supporting vector quantity description and taking the impact of new features to models into consideration,this paper puts forward a new kind of recognition method based on the increment of feature.Thus,improving the ability of identification of new types by expanding the space of model feature.At last,this paper does an empirical research on the complaint text of a business communications company.The experimental result shows that the model and method are effective,and can effectively classify the business complaints.Moreover,it has a stronger recognition ability to new types.