薪資管理一直是人力資源領域中的重要議題,一方面求職者希望為自己獲得優渥待遇,另一方面企業欲使用最少的成本獲得高價值人才,但薪資考量因素難以量化,尚無有效的評估方法,導致雙方對薪資認知差異,因此薪資管理議題需藉由完整的大量資料提供,透過人力銀行大數據分析建構客觀的薪資管理模式,達到更謹慎的管理效果。本研究目的為發展薪資管理資料挖礦與大數據分析架構,整合隨機森林(randon forest, RF)與決策樹(decision tree, DT)技術建立薪資管理模式發掘各職務薪資的影響因素,並與台灣某指標性人力銀行網站合作,進行實證研究,透過歷史數據以平均絕對誤差(mean absolute error, MAE)檢驗方法效度,提升預測準確度。數據評估結果顯示此140類職務準確度平均可有效增加,在四項屬性中最高學歷提升為8.67%、工作地提升為8.30%、公司規模提升為9.08%、產業別提升為9.61%,整體提升準確度8.92%。
Compensation management has been an important topic in the field of human resources. Job seekers want to get generous benefits for themselves, on the other hand companies want to use the least amount of cost to obtain high-value talents. However, the considerations of compensation are difficult to quantify. There are currently no effective evaluation method about pay level. It leads cognitive differences between job seekers and companies. Therefore, compensation forecast needs complete a lot of information provided. Through the Job Bank Big Data analytics, construct an objective compensation forecast model can achieve more discreet prediction effect. The study aims to develop a data mining and Big Data analytics framework for compensation forecast. It integrates random forest and decision tree technology and constructs a compensation management models to explore each job impact factors of the compensation. The study cooperates with a Taiwanese indicative Job Bank Web site for empirical research. Through historical data mean absolute error validates the method validity and improve forecast accuracy. The 140 jobs average forecast accuracy can increase effectively. Highest education level promoted 8.67%, job position promoted 8.30%, comapany size promoted 9.08% and industry category promoted 9.61%. It enhances the entirety predictive accuracy 8.92%.