基層員工晉升預估方法之建構

Abstract

目前國內大部分企業在處理員工晉升問題時一直沒有一套標準的作業流程或有效的科學方法，此外，由於公司之基層員工是組織任務和決策的直接執行者，其對組織變革的績效和成本有很大的影響，且基層員工人數也是公司所有階層員工最多的一層，要決定基層員工晉升人員，除需耗費大量的時間、人力及金錢外，也常衍生出一些人事的問題。因此本研究的主要目的是利用企業現有的內、外部因素及基層員工個人資料，從中找出適用的指標，建構一套能預測員工晉升與否之模式，使企業能夠快速且正確的決定適當之晉升人員以避免員工因不當的晉升而引發問題。此外，本研究亦利用此模式分析晉升及不晉升員工具所具有的特質。本研究方法主要分為四個階段：(1) 選擇變數及收集資料；(2)非量化變數之轉換；(3)分別利用最近鄰居判斷法則（K-Nearest Neighbor rule，KNN）、判別分析（Discriminant Analysis）及倒傳遞神經網路（Back Propagation Neural Networks，BPNN）等三方法構建基層員工晉升預測模式並比較三者之判斷正確率；(4)利用多變量變異數分析（Multivariate Analysis of Variance，MANOVA）找出晉升與不晉升人員間主要之特性差異。本研究最後利用台灣新竹科學園區某公司所提供的員工晉升歷史數據，驗證了本研究方法確定有效可行。

Most Taiwan’s enterprises do not have an effective method to deal with the employees’ promotion problems. Because the first-level employees usually has a large percentage in total employees of an organization and they are also the direct performers of company’s policy. Therefore, it not only requires a lot of time, manpower and money but also involves personnel matters when dealing with the first-level employees’ promotion problems. The main objective of this study is to find appropriate indices from external and internal factors of a company and use these indices to establish a prediction model for first-level employees’ promotion. This prediction model can help enterprises to promote appropriate employees effectively and accurately. The model can also be utilized to find out the major difference between promoted and unprompted first-level employees. The proposed procedure is constructed through five stages: (1) selecting variables and collecting data; (2) transferring non-quantification variable; (3) constructing a prediction model for promoting the first-level employees using K-Nearest Neighbor rule, Discriminant Analysis and Back Propagation Neural Networks respectively, and comparing the effectiveness of these methods; (4) finding out the major difference between the promoted and the unprompted employees using the multivariate analysis of variance method. Finally, a real case provided by a Taiwan’s IC company is utilized to demonstrate the effectiveness of the proposed procedure.