Application of Neural Network On The Taiwan Training QualiSystem Impetus of Performance Evaluation

• Main Authors: Shih-Wei Hsu, 許世葦
• Other Authors: Wen-Tsann Lin
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/56781696644804933864

碩士 === 國立勤益科技大學 === 工業工程與管理系 === 99 === This research analyzes and explores database of Bureau of Employment and Vocational Training (BEVT) on enterprises applying for Taiwan TrainingQuali System (TTQS). First, from the analysis of variance (ANOVA) the results can be obtained that the "assessments" for the more significant differences in the project, and as a research and analysis goals (variables) of the selected variables and thus only effective features selected and remove the problems associated with less features, the classification results to improve the accuracy. Analyzed by the difference for five groups as Platinum medal, Gold medal, Silver medal, Bronze medal and Threshold to find out the differences among these groups the assessment of most indicators and view the results of the classification accuracy. In order to improve the prediction on the Back-Propagation Network (BPN) and the correcting rate, so by Discriminant Analysis and Feature Selection to inspect the importance of "assessment results" and the "assessment indicators". Major component analysis in the former places to view the components of the number of factors should be extracted, the latter are "assessment results" as the goal, and to view "assessment indicators" for "assessment results" has a more important feature of the project selection. To ensure the “Discriminant Analysis“ and the “Feature Selection" of the evaluation selection tool for analysis, the use of "Correlation Analysis" the characteristics of different tools for analysis of assessment indicators capture the extent of differences between projects and to assess whether there is correlation among indicators of the project. Then, data mining techniques in the back-propagation neural network classification techniques to assess TTQS database of the best network architecture and performance. Finally, the two-stage clustering method is "Self-Organizing Maps (SOM) and K-means" to verify the nature of the comparison, comparing the level of verification assessment of the consistency of clustering results, expected to identify the impact the effectiveness of the organization in carrying out TTQS factors to enhance human resources training and effective strategies.