Test Project — Algorithmic Trading Prediction-based Trading & Event-based Backtesting

Implement a class that uses event-based backtesting to backtest the following prediction-based strategy:

Data from http://hilpisch.com/ref_eikon_eod_data.csv. Select one symbol from the data set. Create the following features: log return direction (up or down) log return as 5 categories two SMAs (short and long window) difference between the SMAs two EWMAs (short and long window) difference between the EWMAs two rolling volatilities (short and long window) Split the data set into training (70%) and testing data. Normalize the training features data to have zero mean and standard deviation of one. Normalize the test features data by the same moment values as the training data. Create lagged features data for 5 lags. Train and (back-)test the following algorithms for directional (long/short) trading (from scikit-learn): GaussianNB() LogisticRegression() DecisionTreeClassifier() SVC() MLPClassifier() Compare the performance of the different models numerically.