傳記式資料過去多用來預測員工的任務績效，也獲得了很多不錯的結果，然而預測員工的任務績效於對今日的組織而言，已經無法滿足組織利害關係人多樣的需求，因為好的員工除了自己份內工作需要有好表現之外(任務績效)，也要展現組織公民行為(情境績效)，更要隨組織策略的更迭而有隨之改變(適應性績效)，因此就一個好的甄選工具而言，這三種效標都要加以關心，才能補足理論的缺口與實務的需求。另外，員工離職對組織有相當負面的影響，學者建議應多以有效甄選的方式來降低離職率，因為運用甄選的方式來降低離職率對企業的成本是最低的，他們也建議傳記式資料為可運用的工具之一。經運用個案公司櫃檯人員之傳記式資料及工作表現資料分析，並控制共同方法變異的問題後，發現傳記式資料與員工之任務績效、適應績效、情境績效與離職傾向都有顯著的關係，證明傳記式資料對員工行為擁有良好的預測力，是優秀的甄選工具。 企業人員甄選過去多以線性模式去呈現預測值與效標的關係。這些模式假設預測值上極端的分數總是最佳的。然而許多的研究已發現人格變數與工作績效是曲線性的關係，也發現能力與人格特質間或能力間有所謂互動效果；假若資料型態是非線性的，且變數與變數間有某種程度的互動效果，則以傳統線性模型加以預測與解釋可能並不適合，必須採用新的數學模式方能展現出甄選工具的解釋力與預測力，本研究認為類神經網路是可能的途徑之一，經與迴歸模式比較後發現，於六種工作指標中，類神經網路於五種工作指標之準確率較迴歸模式來的佳，證明類神經網路是未來可考慮採用的工具之一。 從1990年代起，許多的學者提出運用專家系統於人力資源管理之決策上，然而到目前為止，多僅是概念性的陳述，國內外真正應用專家系統於人員決策上仍屬少見，在這少數的研究中，絕大多是開發出雛形系統供研究使用，實作上成功的專家系統則更少。本研究運用利用傳記式資料及員工的績效表現，訓練類神經網路，建立智慧型決策支援系統，協助管理者判斷候選人適任程度，系統架構、介面與資料分析方式可提供研究者及人資工作者參考。 In the past, the biodata have been used to predict task performance. However, the selection instrument that merely predicts task performance is not enough to meet the organization stakeholders’ needs. The employee should be evaluated by whether he/she performs his/her job well (task performance), exhibits organizational citizenship behaviors (contextual performance), and is ready to make a change after the organizational strategy altered. The selection instrument should be built with an emphasis on these three criteria. Since higher turnover rate has a negative effect on organization, scholars have suggested that organizations should reduce costs due to the turnover through selection. Researchers have recommended the use of biodata as an alternative for selection. By controlling the common method variance, this study found that biodata has a significant relationship with task performance, adaptive performance, contextual performance and employee intention for turnover. This research demonstrated that the biodata can be a good selection tool to predict employees’ work behaviors. Decisions made by organization on employee selection most often used the linear model to demonstrate the relationship between assessed factors and criteria. The linear mode assumes that candidates who have extremely high scores in the independent variables are the best ones. However, relevant research has found a curve linear relationship between personality and job performance. It also discovered that there are interactions among personal traits. Those data that demonstrate non-linear and interaction relationships between the variables are not adequate to use linear approach to interpret selection model. In other words, using a non-linear method such as neural network can be a possible solution. After comparing the result of regression and neural network approach; the neural network outperforms regression approach model in 5 predicting criteria. The neural network approach could be a good statistic method in future selection studies. Since 1990, many scholars have proposed that expert systems could be used in human resource management decision making. However, most of the expert system research has simply revealed conceptual framework and only few studies developed the prototype system. Among these few, little research has developed expert systems which can be applied to the actual environment. This study used biodata and employee performance data to train the neural network and use the trained neural network to build the intelligent decision support system. This system can assist HR to evaluate the candidate potential. The architecture, user interface and data analysis method of the system could be a good reference for researchers and HR practitioners.
