隨著科技日新月異的進步,利用智慧數據分析方法來預測股市的現象也越趨頻繁,相較於傳統的時間序列分析與預測,現行以人工智慧做為模型的預測方式更為準確,方法也更為多元。 本研究主要是藉由支援向量機(SVM)、決策樹(DT)和極限梯度提升(XGBoost)演算法,來預測聯電、台積電以及世界的股價漲跌,並比較各模型的預測效果,以及整理出最能影響其預測效果之技術指標。 經本研究實證結果顯示,在預測半導體業的股價上以支援向量機(SVM)最佳,其準確率都高達80%以上,並且發現時間範圍從2016年01月01日至2020年12月31日的短期範圍(5年)預測會受到新型冠狀病毒(COVID-19)影響,但從時間範圍2011年01月01日至2020年12月31日的長期範圍(10年)來看,可以發現股價部會受到影響,準確率皆有平均80%以上的水準,以及發現隨機指標(K)、隨機指標(D)、12日威廉指標(W%R12)、買賣力指標(AD)、5日移動平均線(MA5)、12日相對強弱指標(RSI12)、6日乖離率(BIAS6)以及6日心理線(PSY6),共8項不相同的技術指標為預測半導體業股價的參考指標。
With the rapid progress of science and technology, the phenomenon of using intelligent data analysis methods to predict the stock market has become more frequent. Compared with the traditional time series analysis and prediction, the current prediction method using artificial intelligence as the model is more accurate and the method is more diversified. This research mainly uses the support vector machine (SVM), decision tree (DT) and extreme gradient boosting (XGBoost) algorithms to predict the trend of the stock prices of CMC, TSMC and VIS, and compares the prediction effects of each model, and sorts out the technical indexes that can most influence the prediction effect. The empirical results of this research show that the support vector machine (SVM) is the best in predicting the stock price of the semiconductor industry, with an accuracy rate of over 80%, and found that the time range from 2016 jan 01 solstice December 31, 2020, short range (5 years) prediction will be will be coronavirus (COVID - 19), but from the time range 2011 jan 01 solstice December 31, 2020 (10 years), long range can be found that stock prices would be affected, accuracy is more than 80% of the country's average level, and find random index (K), random index (D), 12, William index (W % R12), sales force index (AD), the 5 moving average relative strength index (MA5), 12 (RSI12), 6 good rate (BIAS6) and 6 line (PSY6), a total of eight different specifications for reference predictor of semiconductor industry.