| 摘要: | 薪資漲幅跟不上通貨膨脹率，如何利用投資增加收入對抗通貨膨脹已經是全民重要課題。由於台灣屬淺碟型經濟體，對海外市場需求高，缺乏吸收震盪的能力，就如同淺碟子般，只要市場上有些微風險即會產生經濟波動，並會反映在股市上，且投資者在選擇投資標的時往往只有對各別產業進行分析，不論基本面或技術面分析皆也只能針對單一個股，忽略產業的獲利表現可能會受到市場風險的影響，且各種產業受景氣的影響程度不同，可能導致在同個景氣階段下營利表現也不同。
景氣指標與股價有高度相關，因此本研究以景氣指標和PMI、NMI採購經理人指數構成元素作為預測指標，並搭配個股基本面、技術面及籌碼面指標，減少個股受整體景氣影響導致誤判，進而彌補基本面、技術面與籌碼面分析的不足，再針對各產業分別用該產業的指標，以台灣股市為標的，並同時帶入AdaBoost (Adaptive Boosting)、人工神經網路 (Artificial Neural Network, ANN)、決策樹 (Decision tree, DT)、梯度增強 (Gradient Boosting, GB)、K-近鄰演算法 (K-nearest Neighbors, kNN)、邏輯式迴歸 (Logistic Regression, LR)、單純貝氏分類器 (Navie Bayes, NB)、隨機森林 (Random Forest, RF)、隨機梯度下降法 (Stochastic Gradient Descent, SGD)及支援向量機 (Support Vector Machine, SVM)等十種機器學習方法建立預測模型，以有效推估未來的趨勢，作為台灣股票投資者參考。;According to the high inflation rate, an efficient investment seems to be a great issue nowadays, especially in countries with thin market like Taiwan. This paper constructed predictive models by using ten kind machine learning method with Economic Indicators for stock price trend prediction.
In this paper, we investigated and done experiments to verify the correlation between Business Indicators and stock price. Revealing that there are three kind Economic Indicators highly correlate with stock price which are Business Indicators, Purchasing Managers’Index (PMI), and Non-manufacturing Purchasing Managers′ Index (NMI). We predicted Taiwan’s stock price trend by using 10 machine learning method, including Adaptive Boosting, Artificial Neural Network (ANN), Decision Tree (DT), Gradient Boosting (GB), K-nearest Neighbors (kNN), Logistic Regression (LR), Navie Bayes (NB), Random Forest (RF), Stochastic gradient descent (SGD) and Support Vector Machine (SVM) with the Business Indicators mentioned above. |
