| 摘要: | 員工職涯發展是一個終身發展與成長的連續序列，人在一生的職涯旅程中，會經歷各種職業或工作，並且在其中扮演某種職務角色、承擔責任或履行義務，人進入職場後便努力地爬升至頂端；然而在經濟環境變遷下，企業經受組織整併與再造、員工忠誠度下降、全球化人才競爭等衝擊，員工職涯發展的思維興起，企業也期許員工個人職涯能與組織目標相契合，以創造更大的組織績效。基於前述的觀點，在過去評估員工晉升之相關研究中，通常把員工的工作績效表現做為預測其未來晉升可能性之工具，而在探討影響員工晉升的因素時，亦把工作績效表現視為主要決定因素之一，但現今企業已不僅評估員工的績效來提升其職位，是轉而將其職涯規劃與發展狀況納入晉升決策之考量，因此，本研究主要目的是透過資料探勘技術中的決策樹演算法進行資料分析，探討影響員工職涯發展與晉升之關鍵因素，並建立員工晉升的預測模型。
本研究採用個案公司累積12年之人力資源發展相關資料為研究標的，進行決策樹C4.5演算法之資料分類與分析，有效樣本數共計1,344筆，投入變數49個，目標變數為晉升、晉等、晉級3個，最終產出決策樹模型與決策規則集；其中訓練資料集占80%（1,076筆），用於建構與訓練預測模型，其餘20%測試資料集（268筆）則用以測試模型精準度。決策樹模型之建置分為二個階段，第一階段將49個投入變數全部放入預測模型中，歸納出的關鍵因素包含整體績效表現、關鍵人才、總特休時數及參與導師計畫，代表第一階段之三個決策樹模型所導出的關鍵因素與個案公司評估員工晉升之因素相符合；第二階段則刪除與晉升直接相關之投入變數（如整體績效表現、潛力發展與關鍵人才等），三個模型萃取出的關鍵因素包含年資、儲備幹部管理訓練時數、專業訓練時數、參與導師計畫等，這些因素多與人才發展相關，可推論個案公司重視員工向上提升前的知識、技能與管理成熟度的準備。二個階段建模產出的六個預測模型之正確度介於74%～89%、ROC曲線下的面積AUC值介於0.720～0.886，顯示預測模型的預測力與鑑別力具有一定水準。
依據本研究結果，建議企業建立系統化的職涯發展路徑，需考量的因素不只是員工的績效表現與發展潛力，還需評估其專業職能、領導與管理能力、人格特質與職務歷練等各項條件是否皆已符合甚至超越下一個職位或職等所需，方能適才適所。同時，建議後續研究可累積更多樣本資料，不斷地投入模型中進行驗證與修正，以提升預測模型之鑑別力；針對不同職務體系、職位高低、職位異動頻率與速度等亦可進一步探勘，瞭解更多影響員工職涯發展與晉升之原因，以期能站在更務實的角度來衡量員工職涯發展之策略與方案。;Career development of employees is a continuous process in humankind’s lifelong development and growth. During the career journey of the employees, they will experience various occupations or jobs, in which they play a certain role, take their responsibilities or fulfill their obligations. When employees enter into the job market, they will devote to being promoted. However, under the transformation of the economic environment, enterprises have experienced the impacts of organizational intergration and reorganization, the decline of employee loyalty, as well as global talent competition, etc.. All of these reasons has result in driving the employees to consider their career developments. On the other hand, enterprises will expect the personal career development of their employees to fit with the organizational goals, which will lead to a better organizational performance. Based on the aforementioned perspective, previous studies regarding employee promotion adopted the employee performance as a tool to predict their possibilities of future promotion. When it comes to the factors of employees’ promotion, work performance is regarded as one of the main determinants. In the practice, however, firms are no longer evaluating work performance for promotion; instead, firms consider the employees’ longterm career planning currently when conducting promotion decisions, and further, develop a complete talent development program. Therefore, the main purpose of this study is to explore the key factors that influence career development and promotion via the decision tree algorithm in data mining, and further to establish a prediction model of employee promotion.
This study adopted the personnel data of the case enterprise as the research sample and conducted a C4.5 decision tree algorithm to classify and analyze the data. There were 1,344 valid samples with 49 input variables. The target variables were promotion, promotion of rank and promotion of sub-rank. The final output is the decision tree model and decision rule-sets. Of the valid samples, 80% were training data (1,076), which was used for building and training prediction models, and 20% were test data (268), which was used for testing the accuracy of the model. There were two phrases of the decision tree model building. The first phrase was to input all 49 variables into the prediction model, and conclude the key factors, including the overall performance, key talents, total annual leave hours and participation in mentoring program, which representing that the three key factors of the first phrase derived from the decision tree model was consistent with the promotion evaluation factors of the case company. As for the second phrase, the input variables directly related to promotion (such as performance evaluation, potential development and key talent etc.) are removed. The result of the extraction showed that the key factors, including tenure, reserved supervisor training hours, professional training hours, participation in mentoring program, etc. were related to talent development. It is possible to imply that the case company pays a significant attention on the preparation of the knowledge, skills and maturity before the promotion of employees. The accuracy of the six prediction models ranged from 74% to 89%, and the AUC value under the area of ROC curve ranged from 0.720 to 0.886, indicating that the predictability and discrimination of the prediction models were at a certain level.
According to the findings of this study, it is suggested that the enterprises should establish a systematic career development path. Firms should consider not only the performance and development potential of the employees, but also evaluate whether their professional capabilities, leadership, personality traits and experience as well as other conditions meet their criteria or even surpass the next position. By doing this, it is possible to put the right employee in the right position. Moreover, it is suggested that the future research could collect more data to continuously input into the model to verify and modify, in order to improve the discrimination of the prediction model. Furthermore, future studies could also explore different job systems, job levels, frequency and speed of promotion, etc. to understand the factors that affect the career development and promotion of employees. This could lead the firms to create strategies and programs to measure employees’ career development from a more pragmatic perspective. |
