隨著經濟成長以及工資和物價的上漲,通膨的壓力造成產銷成本增加,除此之外,台灣在加入WTO之後,所面對貿易自由化的趨勢,必須與全球競爭,而使得以小農經濟為主的台灣面臨重大挑戰,因此能夠取得正確及有用的資訊,並掌握市場供需的變動,才足以因應市場的變化。本研究以台灣民眾日常生活常食用的甘藍作為例子,在農產生成過程中充滿著不確定性的情況下,將生產成本、進出口貿易、產地物價、交易量、氣候等資訊作為可能影響之變數,並利用資料採礦的技術建立CRISP-DM流程,包含迴歸分析、時間序列、類神經網路、SVR、隨機森林、MARS等預測方法找出最佳之農業價量預測模型,在交易量以MARS為最佳模型;在平均價以SVR為最佳模型,研究結果期望能協助相關單位快速取得詳細的產銷預警制度,提早擬定因應措施,達到提升農業資訊力、農產穩定等目的。
As the increase of the economy, wages and consumer price, the pressure of Inflation causes production and marketing cost to increase. In addition, after joined the WTO, Taiwan faced the trend of trade liberalization, and had to compete against the whole world, this situation lead Taiwan to be confronted with a significant challenges. Therefore, to get correct and useful information, and grasp the changes of market supply and demand are able to react the changes in market.This study took cabbage for example, under the situation that full of uncertainty in the process of agricultural produce, regarded import and export trade, origin price, trading volume, and climate information as influence variables, and used data mining techniques to establish CRISP-DM process included regression analysis, time series, neural network, SVR and Random Forests and MARS prediction methods to find out the best agricultural forecasting model of crop price and yield. The results showed that MARS is the best model in Yield and SVR is the best model in price. This study expect the results can assist the related governmental units to obtain detailed price and yield early warning system quickly, and make countermeasure in advance, to improve the ability of agricultural information and production stability.