Research On Application Of Employment Guidance For College Graduates Based On Improved Semi-Supervised Self-Training Method

Pre-employment prediction and analysis is an important tool to effectively link up college enrollment,training and employment,which provides a direction for improving the quality of college students' employment.At present,colleges and universities have established relatively complete students information,which contains valuable information such as students' origin,academic status,employment situation and so on.It can help colleges and universities to dig out the hidden relationship between employment and various factors.With the development of computer information technology,the application of data mining technology in the field of education makes university management more convenient.In this study,the improved Semi-supervised classification method is applied to graduate employment prediction.Because of the difference of situation policies and student group style,the distribution of employment situation of different grades of students can not be fully matched.The Semi-supervised classification method can use graduate samples without employment information to train and expand training set,which can make the prediction model more realistic.This study applies the improved Semi-supervised classification method to the graduate employment forecast and verifies the prediction results.In summary,the main work of this paper is as follows:(1)Analyze and summarize the significance of employment guidance for college graduates and the shortcomings of employment guidance work,and propose to establish a graduate employment forecasting model to improve the efficiency of employment guidance.Secondly,expound the significance of the employment forecasting model of college graduates.Draw the basic link of constructing the forecasting model based on data mining method through analyzing the factors affecting the employment situation of graduates and introducing the data mining method to the employment forecast of college graduates,.(2)Introduce several common semi-supervised classification algorithms systematically,and focus on the semi-supervised self-training classification algorithm.In view of the low accuracy of the self-training algorithm based on Naive Bayes whenthe number of samples is not well distributed,some improvements are proposed.The improved algorithm introduces the similarity calculation method into semi-supervised self-training method.By calculating the similarity between unlabeled samples and marked samples-Euclidean distance and cosine similarity,the samples with high confidence are selected and added to the training set,and then the Naive Bayesian classification model is iterated until the final training is completed.Finally,the effectiveness of the improved algorithm is verified by simulation experiments on UCI data sets.(3)Data collection and preprocessing for graduate information of Chongqing S colleges.In MATLAB,the semi-supervised naive Bayesian self-training combined with similarity algorithm is used to testing the collected data sets,and the prediction accuracy and efficiency are compared of the new algorithm with the other algorithm.In comparison,the improved algorithm has obvious advantages and can better predict unlabeled samples,which provides a reference for the next step of employment guidance.Secondly,the employment forecasting model of college graduates is used to predict the graduates' data,and the results are combined with the actual situation of S colleges to propose employment guidance strategies.