Data

1. Yuanta50
2. Taiwan50
3. orderbook (day orderbook data for INDEX)
4. oil_price
5. bond(fiveyear + ten year)
6. TXF
7. EXF
8. FXF
9. E_F(EXF/FXF)
10. Nikkei
11. SP(S&P 500)
12. DJ(Dow Jones)
13. Russel
14. VIX
15. SOX
16. pcr(put call ratio)
17. FOI(foreign investor OI)
18. IOI(Investment Trust OI)
19. DOI(Dealer OI)
20. inf(infaltion rate)

Model for ETF prediction

Contents of jupyter notebook

• Regression model for Yunta50
• Regression model for 18 ETF
• Classification model for 18 ETF
• Display for .py
• Exploration of Yunta50

Contents of .py

• load.py(return Dataframe)

  • load_stock(fname)
  • load_ETFtable(fname)
  • load_ETFindex(fname,stock)
  • load_ETF(ETFtable,ETFcode)
  • load_orderbook(fname,ETF) : load orderbook and make the date of it as same as the ETF

• price_onedaymodel.py

  • one_day_prediction_18(ETFtable,ETF_list,window,lamb): return ETF_score,ETF_model,ETF_features
  • one_day_prediction(ETF,window,lamb) : return score,model,ETF_features
  • feature_importance(model) :display the weight of every features graph
  • feature_importance_scatter(model,window,ETF_features) : display the weight of ETF_features graph(group by category)

  lamb : the coeficient of the regularization

• price_fivedaymodel.py

  • five_day_prediction_18(ETFtable,ETF_list,window,lamb): return ETF_week_score(total 18 ETF score numpy array)
  • five_day_prediction(ETF,window,lamb) : return week_score(1 ETF score numpy array)
  • show_week_score_18(ETF_score) : display ETF_week_score for the recent weeks
  • show_week_score(score): display week_score for the recent weeks

• class_onedaymodel.py

  • one_day_prediction_regression_18(ETFtable,ETF_list,window,lamb,interval): return 18 ETF score
  • one_day_prediction_regression(ETF,window,lamb,interval) : print average score
  • one_day_prediction_svm_18(ETFtable,ETF_list,window,interval,kernel,C,gamma,coef0) :print score array
  • one_day_prediction_svm(ETF,window,interval,kernel,C,gamma,coef0) :print average train score and test score

  kernel,C,gamma,coef0 : the same as the sklearn.svm

  interval : the range to define rise of fall

• class_fivedaymodel_re.py

  • five_day_prediction_18(ETFtable,ETF_list,window,lamb) :return ETF_week_score
  • five_day_prediction(ETF,window,lamb): return week_score

  Can use show_week_score_18 ,show_week_score from price_fivedaymodel.py to display the score

• submit.py

  • submit_data(ETFtable,ETF_list,window,lamb) : generate Submission.csv