| 摘要: | 但凡製造業多有配合客戶訂單而起伏的基層作業員需求，在辛苦招募的同時如能針對應聘者可能之任職時長預做判斷，並依結果將應聘者分配至適合部門，則可達到「旺季人穩定、淡季人自離」的理想狀況。
參考大量可能影響員工離職與留任之研究後，本研究主要目的為提供一個值得實踐的思路：不追求基層作業員低離職率，而是將基層作業員應聘時所提供資訊以資料探勘中的決策樹演算法進行分類，進行以基本資料做在職時長預測的前導式研究。將應聘者分為：相對不穩定人員(入職半年內離職)、相對穩定人員(入職半年內不離職)，再於招募分配時將應聘人員與用工需求穩定性不同單位做適配，藉此協助企業節省大量人事成本，從源頭達到用人收放自如之效。
本研究以6,000筆同公司、同工作地、同職等、相近工資與工作內容之已離職基層作業員進行8個特徵之決策樹演算法分類。以6,000人中的70%進行訓練後，30%(1,800人)進行檢驗時可進行正確分類人數為1,176人。約65.33%的正確分配率雖不算高，但實務上藉由未來更大量的數據導入，可預期正確分類率將能有所提升。再者，過往在基層作業員招聘上方針多為重量不重質且分發無特定原則。依本研究結果進行分發，進可為每年招募達百萬人公司節省數以千萬計用人成本，退亦無造成損失之可能性。
同樣的研究流程與方法，有機會用於不同國家、地區、產業的基層作業員招募上，利用招聘時所能取得之特徵因素進行分析，可能將因不同的天時、地利、人和產生各異結果，而這也正是企業了解並客製自身所需人員的機會。期望此研究發想能對所有需要大量基層作業員的企業做出貢獻。
;In manufacturing, there are many Front line operator needs fluctuate with customer orders. We try to pre-judgment the possible length of employment tenure before onboard, and assign the candidates to suitable departments according to the results.
Instead of pursuing a low turnover rate of operators, the main purpose of this study is used data mining (Decision tree) to classify the candidates may unstable (leave within six months of entry) or stable (do not leave within half a year of entry). After that we can adapt candidates to BGs with different stability of employment needs. During this way about recruitment allocation, we may save a lot of personnel cost.
In this study, 6,000 employees who have left the same company, the same workplace, the same level, etc., with similar salary and job content, are classified by a decision tree algorithm based on 8 characteristics. Although the correct distribution rate about 65.33% is not high enough, but can be expected that the rate of classification will be improved after import more data in the future.
The same research process and method may be used in different countries, regions and industries, and obtain different result. It may be an opportunity for companies to understand and customize employ they need and help the enterprises that require a lot of basic operators. |
