謝謝,alexliou:分享文中有一建議
We use tanh here since our signals are also in the +/-1 range. The output of the network is a sigmoid function since we want a prediction in the 0..1 range.
1.依此,用tanh還算恰當。
2.文中提到他有 62% prediction accuracy。That point inspires me to produce my own prediction accuracy。
3.另外,本來我也有該用多少筆歷史資料進入training的疑問。20天?60天?文中是100筆,可為參考。
Thanks ! Looking forward for more suggestions!!
感謝alexliou分享文中有許多寶貴觀念,摘出與大家分享。發展智能策略的方法:【Step 1】: The target variable,挑選要預測的目標,我選的是:加權指數漲跌幅%
【Step 2】: The features,selecting the features is critical for success。這點我堅信如此。我們也可以先就歷史資料做feature selection或轉軸(PCA主成份分析、FA因素分析)等多變量分析的處理。這裡,我只是憑經驗和直覺,本相關issues,雖重要但暫列他案處理。
【Step 3】: Preselecting/preprocessing predictors,
小點提到:Determine the correlations between the signals. Remove those with a strong correlation to other signals, since they do not contribute to the information.不要放入高度相關的二個變量。
【Step 4】:Step 4: Select the machine learning algorithm,選擇演算法,在此是BPN(其實NN可能有人不會覺得有用,也許吧?)
小點提到:
The usual method is error backpropagation. But it turns out that the more hidden layers you have, the worse it works. The backpropagated error terms get smaller and smaller from layer to layer, causing the first layers of the net to learn almost nothing. Which means that the predicted result becomes more and more dependent of the random initial state of the weights. 我隱藏層只有一層是OK的,我也覺得過多無益,如上文所述,蠻好的概念。
【Step 5】: Generate a test data set,用EXCEL下載證期交所資料。
【Step 6】: Calibrate the algorithm,還沒到此,等BPN模式確認了,需進一步微調參數等。
【Step 7】: The strategy,漲跌幅%預測出來後,如何操作應該不難。
【Step 8】: The experiment。需要,會的。