應用資料探勘分類技術於人才甄選之研究

在人力資源管理領域中，選才是一切人力資源管理的開始，而如何找到適任的員工也是各成功企業或組織在經營管理上最常面臨的問題之一。而欲做好人才甄選的工作，發展有效的甄選工具是其中的關鍵所在。常見的人事甄選工具包含履歷表、面談、推薦函、自傳、能力測驗及人格測驗等，其中面談是最普遍使用的甄選方法。而企業通知面談之前，第一道手續就是從成千上萬的履歷表中篩選符合職務需求的人才。對於很多企業來說，篩選履歷表是一件相當繁重的工作，且大多數的審核人員往往僅以經驗法則或主觀意識來判斷，所以需要一套方法讓篩選工作變得較客觀且有效。本研究導入資料探勘技術於人才甄選之研究，將利用資料探勘中的分類技術，以履歷表中所有的項目資料為輸入變數建構員工績效的分類模式。並且透過所建構的資料探勘分類模式，並找出履歷表中影響員工績效的重要變數，作為企業往後甄選新進人員的初步篩選準則。本研究採用的資料探勘技術包含多元適應雲狀迴歸（multivariate adaptive regression splines, MARS）、分類迴歸樹（classification and regression tree, CART）、類神經網路（artificial neural networks, ANN）與支援向量機（support vector machine, SVM），並且以某銀行電話行銷人員的一年工作績效以及履歷表資料為實證資料。實證結果顯示，透過MARS與CART技術，本研究找出的影響員工績效的重要履歷表變數分別是「星座」、「婚姻狀況」、「應徵原因」及「電腦能力」等四項。並且，在四個分類模式的鑑別績效比較上，利用BPN所建構的履歷表分類模式能有最高的鑑別正確率，因此利用BPN分類技術可以協助企業從履歷表中篩選出能夠產生較高績效的員工。

關鍵字

資料探勘；人才甄選；分類問題；履歷篩選

並列摘要

The success of a business depends on the recruitment of right people for the right positions. Interview is one of the most popular methods for recruitment. Before interview, the first step is to sift the information from the resume carefully to find right person for every specific position. It is difficult to do the filtering without advisable tools. In the research, we apply data mining techniques for recruitment, especially aim at filtering resumes. The techniques including multivariate adaptive regression splines (MARS), classification and regression tree (CART), backpropagation neural network (BPN), support vector machine (SVM) are adopt to build classification models using every variables in the resume that may influence the performance of employees. Experimental results showed that the four variables including ”Constellations”, ”Marriage”, ”Reason for applying this job” and ”Computer skill” are selected as the most important variables for the recruitment of right people. Moreover, BPN model outperforms the MARS, CART and SVM models in terms of resume filtering and hence provides an efficient alternative to help companies to recruit right employees which produce higher performance.