Research On A Ensemble Learning-based Postgraduate Quality Analysis And Evaluation System

In recent years,Chinses education is continuing to develop,the number of graduate students enrolled in record high,the training process of the decline in the quality of training,academic thesis misconduct and other problems gradually highlighted.At present,the application of machine learning in the evaluation of postgraduate training quality in China is still immature,accompanied by the continuous improvement of the informationization mechanism of universities.The data related to the graduate student training process are accumulating,and if it is analyzed and applied in depth,it can obtain potential information that is difficult to find manually,and help the relevant departments to improve the quality of graduate student training to a certain extent.In this thesis,the data related to postgraduate training are analyzed in this context,and a machine learning-based postgraduate training quality analysis and evaluation system is designed and implemented.The main work contents are as follows.(1)Analyzing and pre-processing data of graduate student training process.Firstly,the original dataset is constructed by analyzing and integrating the graduate training data.Secondly,the original dataset is pre-processed to avoid the impact of missing values and outliers on the model,while irrelevant features and redundant features in the dataset are eliminated to obtain a subset of feature attributes that optimize the classification performance of the model,and normalized by Max-Min standardization to construct the graduate training dataset.(2)The graduate student training quality evaluation model is constructed.To address the problems of imbalance and multiple classifications in graduate training quality evaluation,KSRS,a graduate training quality evaluation algorithm fused with K-means SMOTE and Random＿Stacking is proposed.Firstly,the K-means SMOTE algorithm is used to adjust the sample distribution of the dataset to make it approximately balanced in each category.Secondly,multiple machine learning algorithm models are used for parameter search optimization using grid search method,and four models suitable for the scenario of this thesis are selected for combination,and two fusion models are modeled separately using Random＿Stacking algorithm.Finally,the advantages of KSRS models for graduate training quality evaluation are demonstrated by experimentally comparing two KSRS fusion models,two Stacking fusion models,Cat Boost and Ada Boost.(3)The graduate student training quality analysis and evaluation system is developed.According to the above machine learning model,combined with the current demand for the cultivation quality system,Vue.js,Echarts,Python and other technologies are used to build the postgraduate training quality analysis and evaluation system.The system realizes the multi-angle analysis of postgraduate cultivation data based on visualization technology and the evaluation of postgraduate cultivation quality based on machine learning model,and conducts functional and performance testing after the initial completion of the system to ensure system integrity and stability.