M_q - arcturus9/useful-link GitHub Wiki

Data Analysis

• statistics for each stock code ==> compare (mean / stdev) of Y for two stock code..

Data Preprocessing

• (MinMax, Standard) Scaler ==> X_train_s2, s3, s4, ...
• statistics for each stock code ==> Scaler for each stock code ==> X_train_ScaleByCode1
• Generate 2day, 3day set : {[D-0] + [D-1]}, {[D-0] + [D-1] + [D-2]}, ... ==> X_train_2days, X_train_3days

Selecting Baseline Model

• Cross validation (ExtraTreesClassifier)
• Model selection among ExtraTreesClassifier, AdaBoost, Gradient Boosting Method, RandomForests (using GridSearch, using EstimatorHelper)
• Hyper-parameter Optimization for Best Model

Compare Different X_train_xx... for selected Baseline Models

• X_train_s2, X_train_ScaleByCode1, X_train_2days...

Deep learning

• RNN
• LSTM
• BERT

Ensemble DNN + (AdaBoost, GBM)

• DNN
• Adaboost, GBM ...