應用灰色系統理論與類神經網路於農產品銷售量預測之研究

Study of Agricultural Product Sales Forecast with Grey System Theory and Artificial Neural Networks

鄭米茹(Mi-Ru Cheng)

指導教授 ： 王永心

大同大學/經營設計學院/資訊經營學系所/碩士(2005年)

摘要

隨著台灣加入WTO，農產品的銷售競爭，不再侷限於國內，而是邁向了國際化，導致銷售量呈現不穩定的現象。一個以營利為目的的企業組織，產品銷售量的多寡將影響其發展與生存。預測具有展望未來的功能，預測的結果往往是管理者的決策依據，故預測的準確度影響企業的發展，所以預測對企業而言，扮演著極重要的角色。為了降低風險，確實掌握市場脈動，傳統農業經營者不能只憑經驗來預測產品銷售量。因此，本研究主要利用灰預測與倒傳遞類神經網路二種預測方法，以雲林縣某蔬菜供銷中盤商做為研究個案，預測農產品銷售量，以「蔥」為例，實證比較分析兩種模型何者精確度為佳，根據預測結果，提供廠商經營上之採購等作業的決策參考。研究結果顯示，以預測精確度評量來看，BPN的預測誤差低於5%，屬於「優」的等級；灰預測的預測誤差介於5%~10%之間，屬於「良」的等級，平均而言，兩者的精確度都高達九成以上，預測能力皆有一定之水準。

關鍵字

灰色系統理論 ； 類神經網路

並列摘要

Since Taiwan joined the WTO, agricultural product sales is no longer limited to domestic competition but steps to the internationalization. Currently, the sales volume presents an unstable phenomenon, which affects the development and survival of an enterprise that is profit-based. The prediction has the function for forecasting future and the result is often the regulatory authority decision-making basis. Therefore, prediction accuracy influences enterprise's development and plays extremely an important role in the enterprise. The traditional agriculture operators should not only depend on experiences to predict the product sales in order to reduce the risk and truly grasp the market pulsation. Therefore, this study uses the Grey Prediction and the Back Propagation Network in dealing with forecast problems. We use an example of the vegetable middle distributor industry’s supply and marketing situation of Yunlin County as the research case to forecast agricultural product on market sales volume using “scallion” as example. Analysis that makes use of the accuracy of the two models according to their forecast results will provide the factory owner with an effective consultation on matters pertaining to procurement and business operations. The results regarding forecast accuracy measures show that BPN’s error percentage is less than 5%, which is classified as “excellent”, and Grey Prediction’s, on the other hand, has a percentage between 5%~10%, which then is under the “good” class. These two forecast methods are seen to yield considerably accurate results, and are most likely to reach high forecasting quality.

並列關鍵字

Arificial Neural Networks ； Grey System Theory

參考文獻

25. 李宗儒、鄭卉方，「應用灰色理論預測農作物之價格—以紅豆為例」，農林學報，49(2)，P83-92，2000年。

40. 台灣省政府農林廳，台灣地區農產品批發市場年報，2000~2003年。

1. Al-Saba, T. and El-Amin, I., “Artificial neural networks as applied to long-term demand forecsting,” Artificial Intelligence in Engineering, Vol. 13, pp.189-197, 1999.

2. An, S., Yan, J. and Yu, X., “Grey-system studies on agricultural ecoengineering in the Taihu Lake area, Jiangsu, China,” Ecological Engineering, Vol. 7, pp. 235-245, 1996.

4. Box, G.E.P. and Jenkins, G.W., Time Series Analysis, Forecasting and Control, Revised Edition, Holden-day, 1970.