應用資料採礦技術於大學生就業發展方向之研究

Application Study in the Development of University Students' Employment by using Data Mining Technology

王智立(Chih-Li Wang) ； 陳克琛(Ker-Chang Chen) ； 謝孟仁(Meng-Ren Hsieh)

《智慧科技與應用統計學報》 17卷2期 (2020/01) Pp. 17-39

摘要

隨著資訊科技迅速的發展大量資料的累積已相當普及，但要從巨量數據（big data）中尋找出有用的資訊並不簡單，該如何挖掘出有價值的資訊給決策者，做出有效的決策，這是值得去探討的。資料探勘（data mining）是近年來逐漸蓬勃發展的一項技術，是一種結合數個不同問題領域（problem domain）的專業技術，並可找出資料中資訊的一個流程。本研究以大專校院就業職能平台（Career & Competency Assessement Network, UCAN）實際資料中的16類和畢業生畢業後一年就業調查資料為例，本研究之資料係以學生在校期間（大學一年級）所進行之UCAN職涯興趣測評結果，與該學生畢業後一年實際之就業類型，進行預測模式之建構，並對此模式進行預測準確率評估，因此，本研究之受測者資料係連貫4-5年的縱貫性資料（Longitudinal data）研究。另外，本研究再結合Holland（1985）理論的6大類職涯類型，找出學生測評結果與學生畢業後工作職場之一致性。最後，本研究利用類神經網路（Neural Network）及邏輯斯迴歸（Logistic Regression）資料探勘技術模擬分析，結果發現UCAN的16類職涯類型在學期間之測評結果與畢業後一年的實際就業類型在兩種方法建模準確度從8.57%分別提升到82.58%和97.05%；而結合Holland（1985）6類型測評結果與畢業後一年的實際就業類型的兩模型準確率也從22.70%分別提升到99.31%和99.58%。因此，可以發現類神經網路和邏輯斯迴歸兩種資料採礦方法之預測結果皆有不錯的表現。

關鍵字

資料探勘 ； 大專院校就業職能平台UCAN ； Holland碼 ； 類神經網路 ； 邏輯斯迴歸

並列摘要

With the rapid development of information technology, a large accumulation of information has become quite common, but finding out useful information from huge amount of data was not easy. It is worth exploring how to dig out valuable information to make effective decision-making. In recent year, Data-mining is a professional technique which was gradually developing and combining several different problem domain. In this study, we were use the UCAN database which divided into 16 career types to build a model to predict the career types after graduation. In addition, we combine with the John Holland’s theory (1985) which divided into 6 career types. We use two kinds of data mining technology to build models. One is Neural Network, the other is Logistic Regression. The data of this study was a coherent 4-5 year longitudinal data. Find out the consistency between student assessment results in school and the working categories after graduation. In result, this study uses Neural Network and Logistic Regression technology to simulation and analysis the graduate employment information data. we build Neural Network model and Logistic Regression model to analysis. Finally, we find that no matter use Neural Network model or Logistic Regression model has improved accuracy rate. (1) In the UCAN 16 career types method, the accuracy of Neural Network model and Logistic Regression model have increased from 8.57% to 82.58% and 97.05% respectively. (2) In the Holland 6 career types method, the accuracy of Neural Network model and Logistic Regression model have increased from 22.70% to 93.10% and 99.58% respectively. Consequently, it can be found that the prediction results of both the Neural Network and the Logistic Regression method have good performance.

並列關鍵字

Data mining ； UCAN ； Holland code ； Neural Network ； Logistic Regression