筆記型電腦維修中心的維修品質與速度,是目前消費者選購筆記型電腦品牌時其中一項指標。而筆記型電腦維修零件預測的準確度絕對是影響筆記型電腦維修速度的主因。維修零件的需求一般來說屬於間歇性需求,間歇性需求通常是隨機的,並且有絕大部分為零的需求,由於在非零的需求之間可能會有很大的變異性,因此,在預測間歇性需求時,較難尋找到特定的規律。 因此,本研究應用倒傳遞類神經網路來建立筆記型電腦維修零件預測模式,並與一般常用的移動平均法、指數平滑法及灰預測進行比較。透過誤差均方根做為衡量指標,進行預測之績效評估。 實證研究結果顯示倒傳遞類神經網路的確在預測績效有十分傑出的表現。其次是指數平滑法,而灰預測在遇到大多數為零的需求時,預測績效表現則十分不理想。
The repair quality and efficiency of notebook repair center are the key factors for consumers to choose notebook brand. Notebook spare parts forecasting accuracy is definitely the reason of influencing notebook repair speed. Spare parts are usually belonging to intermittent demand. Intermittent demand is usually random, and most part of its demand is zero. Because of demand of non-zero may exist tremendous variability, hence it's hard to find a specific regular pattern for intermittent demand forecasting. Therefore, this study applied Back-propagation Neural Network (BPN) to build up a demand forecasting pattern of notebook spare parts. The results of demand forecasting pattern will be compared with those of the method generally using like Moving Average, Exponential Smoothing, and Grey Prediction. Using root mean square error (RMSE) to be a key index, and it will proceed the performance evaluation of forecasting. The result of experiments shows that the most excellent performance of forecasting is BPN, and the next is Exponential Smoothing. Grey Prediction using on forecasting the bulk of zero demand is unsatisfactory performance.