Backtesting with Machine Learning

Available Patterns

Only bullish candlestick patterns are available right now and they are:

• Inverted Hammer
• Hammer
• Bullish Engulfing
• Bullish Harami
• Morning Star
• Morning Star Doji
• Piercing Pattern
• Dragon Fly Doji

This software backtest a trading strategy and runs the results through a user defined machine learning, with feature engineering to optimize the strategy as much as possible. An example output of this software would look like this.

How to use

Install all dependencies.

pip install ml-backtest

First you need to import all the required classes.

from ml_backtest import Backtest, MachineLearning
from ml_backtest.machine_learning import CandleStickDataProcessing
from ml_backtest.strategies import InvertedHammer
from ml_backtest.models import RandomForestRegressorTrainer
import pandas as pd

After that make sure you rename your data-frame columns to these names if they are not already named that.

df = pd.read_csv('YOUR FILE NAME.csv')
df = df.rename(columns={'Time': 'date', 'Open': 'open', 'Close': 'close', 'High': 'high', 'Low': 'low'})

After you have your data-frame prepped you can make an instance of the strategy you want and pass it into a Backtest instance.

strategy = InvertedHammer()
backtest = Backtest(df, strategy)
print(backtest.get_results())

After that, we have all the data we need for machine learning to take place. Just declare an instance of the machine learning class and pass the need info into it. The machine learning takes place when the run function is called on the class. We can dump the model (saving the model to be used as a standalone file) with the dump_model function.

ml = MachineLearning(ml_class=RandomForestRegressorTrainer,
                    df=df,
                    results=backtest.get_trades(),
                    rows=10,
                    columns=['EMA_Diff', 'SMA_Diff', 'MACD_hist'])
ml.run(dp_pattern=CandleStickDataProcessing.calculate_inverted_hammer_features)
ml.dump_model(filename='YOUR FILE NAME')

After we trained the model, we want to backtest the model and see the results! The fun part! Two importnant functions you need to call for the backtest, get_util function and get_data function. The get_util will return a tuple of important values to be passed into the backtest class. The get_data will just be the data-frame as before but it includes all necessary added features during the call of feature_engineering function of the desired machine learning class.

model, columns, rows = ml.get_util()
data = ml.get_data()

ml_backtest = Backtest(data, strategy, model=model, columns=columns, rows=rows, cs_pattern=True)
print(ml_backtest.get_results())

Thats it! You should see similar output text wise as the outputs provided above. A more in depth how to use guide to customize your machine learning and strategy can be found below.

• Machine Learning Class Guide
• Trading Strategy Class Guide