近年來,全球對於能源與替代性能源的需求快速增加,然而在多種替代性能源之中,太陽能是未來最具有發展潛力的替代性能源,因此太陽能未來的發展趨勢也備受各界重視。 隨著資訊科技快速發展,財務指標結合人工智慧在預測應用上更為廣泛;其中人工智慧在預測股價指數方面有一定的準確性與可信度,因此希望應用人工智慧預測全球太陽能指數與自行編制的台灣太陽能指數,以了解其未來趨勢。 本研究欲探討三種不同的人工智慧方法在預測全球太陽能指數與台灣太陽能指數的預測績效,並比較三種人工智慧方法,以期發現最適合應用於預測太陽能類股指數的方法和模型。 在研究後發現: 1. 支援向量迴歸無論是在預測全球太陽能指數或台灣太陽能指數方面,其預測誤差最低,衡量預測方向之準確度最佳。 2. 支援向量迴歸預測模型中可以發現,當預測全球太陽能指數時,費城油業指數為最重要的判斷指標;而在預測台灣太陽能指數時,S & P500指數與S & P 500能源指數為最重要的觀察指標。 3. 當支援向量迴歸於預測全球太陽能指數時,若使用較長時間的測試期,其預測準確率與衡量預測方向之準確度會比短時間的測試期為優。
Recently, the demand of energy and substitutable energy surges in the whole world. Among various kinds of alternative energies, solar energy is the most potential one. Therefore, the development and the tendency of the solar energy have raised lots of attention from academic and practical fields. Along with the science and information technology developed speedily, financial indicator combining artificial intelligence had been applied widespread. Artificial intelligence has certain accuracy and credibility in predicting the stock price index. Thus, this study attempts to apply artificial intelligence to forecast the trend of global solar energy index and Taiwan solar energy index. This work also compares the performance of prediction of three kinds of artificial intelligence methods on global solar energy index and Taiwan solar energy index, and attempts to find out which method is the best to forecast the solar energy index. The empirical results are summarized below: 1. This study finds that Support Vector Regression method (SVR) has the lowest error of prediction and the best accuracy of the direction on both the Global solar energy index and Taiwan solar index. Backpropagation neural network is second and Adaptive Network-based Fuggy Inference system is third. 2. Empirical finding indicates that the trend of the Phlx Oil Service Sector Index is the most significant observation indicator to predict the solar index. This investigation also finds that S & P 500 index and S & P 500 energy index are the most significant observation indexes to forecast the trend of Taiwan Solar Energy Index. 3. This work also finds that Support Vector Regression method (SVR) has the lower error of prediction and the better accuracy of the direction on Global solar energy index for long-term testing horizon than that for short-term testing period.