Forecasting Agricultural Output Based on Support Vector Regression and Climate Data

• Main Authors: JHENG, TING-ZUO, 鄭庭佐
• Other Authors: LEE, CHIEN-PANG
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/f9zjhx

碩士 === 國立高雄海洋科技大學 === 海事資訊科技研究所 === 105 === Due to the dramatic changes in the global climate in recent years, the incidence of extreme weather events has increased significantly. Many crops have been adversely affected and Taiwan is a high risk region in global climate change. In this thesis, we used staple food rice in Taiwan as an example to establish the rice yield prediction model and forecast the rice yield in the same period of next year. It is expected that the possible production of rice in the same period of the next year will be known and the relevant measures will be made immediately by the government authorities for the insufficiency or excess of rice production. In the past, there have been many studies on crop yield forecasting using different forecasting methods to predict. In this thesis, one type of machine learning algorithm, support vector regression (SVR), is used as the main method to predict rice yield. Finally, the integration of 12 cities of the rice production forecast model and 12 cities of the others rice production forecast model is developed. The results shown that the root mean square error (RMSE) value calculated from the predicted values of the two forecast models and the actual values of the test data is less than 60 and the correlation coefficient (CC) is higher than 0.996. In the training, the RMSE calculated from the predicted values of the two forecast models and the actual values of the training data is very close to the RMSE from calculated in the testing, and thus we can see that the SVR + Bootstrap + GA model has high reliability and high stability.