本研究依據勞保局之「歷年職業災害勞保給付統計年報」資料，應用灰系統理論之灰預測GM（1,1）模型，期望建立一能應用於我國勞工保險職業災害給付千人率預測之模型。職業災害保險總給付、傷病給付及失能給付千人率，五期至十期的滾動建模均可得到非常良好的預測結果，其預測精準度都大於90%以上。職業災害死亡給付千人率的灰色預測準確性雖較差，但大致而言，以低於八期之建模數據仍可至少在合格範圍內。以本研究之預測模型預測職業災害保險給付千人率未來趨勢，總給付、失能給付及死亡給付呈現下降趨勢，只有傷病給付有上升情形。

This study applied GM (1,1) model of grey system theory to establish the forecasting models of the ratio per thousandth for the labor insurance payments of occupational accidents in Taiwan based on the annual statistics of Bureau of Labor Insurance. The ratio per thousandth of the total benefit payment, the benefit payment on injuries or illness, and the benefit payment on disability represent excellent results in the rolling forecast modeling on 5 to 10 periods. As to the benefit payment on death, the accuracies of grey prediction on the ratio per thousandth are generally acceptable on less than 8 periods. Predicting the trend of insurance payments of occupational accidents, the ratio per thousandth is going down on the total payment, disability payment and death payment, and rising on the injuries or illness payment.