Research On Identifying Poverty-stricken College Students Based On Campus Card Consumption Data Mining

With the popularization and development of higher education,more and more impoverished students are able to continue their studies in colleges and universities.However,the students with family financial difficulties can not afford their basic living and learning consumption,because of the soaring prices.Promoted by the need for accurate discriminant of poor students,colleges and universities gradually use written applications to determine and subsidize poor students.However,the number of college students and family situation is too complex to achieve satisfactory results.The widespread use of one-card system in colleges and universities makes it possible to tap students'behavior and consumption habits fully.Therefore,how to extract the effective value information from the miscellaneous consumption data,to realize the classification of students'onsumption,and to provide decision-making basis for university funding work is one of emphasis in the research of poor students'decision-making in colleges and universities.Through the analysis and research on the consumption data of campus cards in a vocational and technical college,the characteristics of students'consumption behavior are excavated to improve decision support for poor students in colleges and universities.The main contents of this paper are as follows:(1)Analyses the forms and components of various consumption data in the current campus card system,summarizes the classification and existing problems of eligibility methods for needy students,and constructs a basic poverty index system,which provides data basis for the realization of feature reduction and judgment of poor students' system in the following chapters.(2)Studies various feature reduction methods widely used nowadays,and put forward the attribute reduction method based on improved principal component analysis.With the help of this method,this paper further deals with the characteristics of the index system of poor students,which reduces the original index to 11.(3)Aims at the problem without obvious category labels of students,this paper takes the special index system as input,uses K-Means clustering method to realize the category labels of students.Using the calculation of students'consumption index as the criterion of poverty discrimination,this paper puts forward a poverty determination model based on GBDT algorithm.The final experimental results show that the new feature reduction method based on improved principal component analysis can effectively reduce the characteristics of the target system while retaining the maximum information.Compared with the classical K-Means algorithm,it can quickly and accurately determine the poverty category of students,which has a good reference significance and decision support for the realization of poor students'decision-making management in colleges and universities.