DataCamp Machine Learning for Finance in Python
MACHINE LEARNING FOR FINANCE IN PYTHON
Modern portfolio theory
(MPT); efficient frontiers
Nathan George
Data Science Professor
DataCamp Machine Learning for Finance in Python
DataCamp Machine Learning for Finance in Python
Joining data
stocks = ['AMD', 'CHK', 'QQQ']
full_df = pd.concat([amd_df, chk_df, qqq_df], axis=1).dropna()
full_df.head()
AMD CHK QQQ
Date
1999-03-10 8.690 0.904417 45.479603
1999-03-11 8.500 0.951617 45.702324
1999-03-12 8.250 0.951617 44.588720
1999-03-15 8.155 0.951617 45.880501
1999-03-16 8.500 0.951617 46.281398
DataCamp Machine Learning for Finance in Python
Calculating returns
# calculate daily returns of stocks
returns_daily = full_df.pct_change()
# resample the full dataframe to monthly timeframe
monthly_df = full_df.resample('BMS').first()
# calculate monthly returns of the stocks
returns_monthly = monthly_df.pct_change().dropna()
print(returns_monthly.tail())
AMD CHK QQQ
Date
2018-01-01 0.023299 0.002445 0.028022
2018-02-01 0.206740 -0.156098 0.059751
2018-03-01 -0.101887 -0.190751 -0.020719
2018-04-02 -0.199160 0.060714 -0.052971
2018-05-01 0.167891 0.003367 0.046749
DataCamp Machine Learning for Finance in Python
Covariances
# daily covariance of stocks (for each monthly period)
covariances = {}
for i in returns_monthly.index:
rtd_idx = returns_daily.index
# mask daily returns for each month (and year) and calculate covariance
mask = (rtd_idx.month == i.month) & (rtd_idx.year == i.year)
covariances[i] = returns_daily[mask].cov()
print(covariances[i])
AMD CHK QQQ
AMD 0.000257 0.000177 0.000068
CHK 0.000177 0.002057 0.000108
QQQ 0.000068 0.000108 0.000051
DataCamp Machine Learning for Finance in Python
Generating portfolio weights
for date in covariances.keys():
cov = covariances[date]
for single_portfolio in range(5000):
weights = np.random.random(3)
weights /= np.sum(weights)
DataCamp Machine Learning for Finance in Python
Calculating returns and volatility
portfolio_returns, portfolio_volatility, portfolio_weights = {}, {}, {}
# get portfolio performances at each month
for date in covariances.keys():
cov = covariances[date]
for single_portfolio in range(5000):
weights = np.random.random(3)
weights /= np.sum(weights)
returns = np.dot(weights, returns_monthly.loc[date])
volatility = np.sqrt(np.dot(weights.T, np.dot(cov, weights)))
portfolio_returns.setdefault(date, []).append(returns)
portfolio_volatility.setdefault(date, []).append(volatility)
portfolio_weights.setdefault(date, []).append(weights)
DataCamp Machine Learning for Finance in Python
Plotting the efficient frontier
date = sorted(covariances.keys())[-1]
# plot efficient frontier
plt.scatter(x=portfolio_volatility[date],
y=portfolio_returns[date],
alpha=0.5)
plt.xlabel('Volatility')
plt.ylabel('Returns')
plt.show()
DataCamp Machine Learning for Finance in Python
MACHINE LEARNING FOR FINANCE IN PYTHON
Calculate MPT portfolios!
DataCamp Machine Learning for Finance in Python
MACHINE LEARNING FOR FINANCE IN PYTHON
Sharpe ratios; features and
targets
Nathan George
Data Science Professor
DataCamp Machine Learning for Finance in Python
DataCamp Machine Learning for Finance in Python
DataCamp Machine Learning for Finance in Python
DataCamp Machine Learning for Finance in Python
Getting our Sharpe ratios
# empty dictionaries for sharpe ratios and best sharpe indexes by date
sharpe_ratio, max_sharpe_idxs = {}, {}
# loop through dates and get sharpe ratio for each portfolio
for date in portfolio_returns.keys():
for i, ret in enumerate(portfolio_returns[date]):
volatility = portfolio_volatility[date][i]
sharpe_ratio.setdefault(date, []).append(ret / volatility)
# get the index of the best sharpe ratio for each date
max_sharpe_idxs[date] = np.argmax(sharpe_ratio[date])
DataCamp Machine Learning for Finance in Python
Create features
# calculate exponentially-weighted moving average of daily returns
ewma_daily = returns_daily.ewm(span=30).mean()
# resample daily returns to first business day of the month
ewma_monthly = ewma_daily.resample('BMS').first()
# shift ewma 1 month forward
ewma_monthly = ewma_monthly.shift(1).dropna()
DataCamp Machine Learning for Finance in Python
Calculate features and targets
targets, features = [], []
# create features from price history and targets as ideal portfolio
for date, ewma in ewma_monthly.iterrows():
# get the index of the best sharpe ratio
best_idx = max_sharpe_idxs[date]
targets.append(portfolio_weights[date][best_idx])
features.append(ewma)
targets = np.array(targets)
features = np.array(features)
DataCamp Machine Learning for Finance in Python
Re-plot efficient frontier
# latest date
date = sorted(covariances.keys())[-1]
cur_returns = portfolio_returns[date]
cur_volatility = portfolio_volatility[date]
plt.scatter(x=cur_volatility,
y=cur_returns,
alpha=0.1,
color='blue')
best_idx = max_sharpe_idxs[date]
plt.scatter(cur_volatility[best_idx],
cur_returns[best_idx],
marker='x',
color='orange')
plt.xlabel('Volatility')
plt.ylabel('Returns')
plt.show()
DataCamp Machine Learning for Finance in Python
DataCamp Machine Learning for Finance in Python
MACHINE LEARNING FOR FINANCE IN PYTHON
Get Sharpe!
DataCamp Machine Learning for Finance in Python
MACHINE LEARNING FOR FINANCE IN PYTHON
Machine learning for MPT
Nathan George
Data Science Professor
DataCamp Machine Learning for Finance in Python
Make train and test sets
# make train and test features
train_size = int(0.8 * features.shape[0])
train_features = features[:train_size]
train_targets = targets[:train_size]
test_features = features[train_size:]
test_targets = targets[train_size:]
print(features.shape)
(230, 3)
DataCamp Machine Learning for Finance in Python
Fit the model
from sklearn.ensemble import RandomForestRegressor
# fit the model and check scores on train and test
rfr = RandomForestRegressor(n_estimators=300, random_state=42)
rfr.fit(train_features, train_targets)
print(rfr.score(train_features, train_targets))
print(rfr.score(test_features, test_targets))
0.8382262317599827
0.09504859048985377
DataCamp Machine Learning for Finance in Python
Evaluate the model's performance
# get predictions from model on train and test
test_predictions = rfr.predict(test_features)
# calculate and plot returns from our RF predictions and the QQQ returns
test_returns = np.sum(returns_monthly.iloc[train_size:] * test_predictions,
axis=1)
plt.plot(test_returns, label='algo')
plt.plot(returns_monthly['QQQ'].iloc[train_size:], label='QQQ')
plt.legend()
plt.show()
DataCamp Machine Learning for Finance in Python
DataCamp Machine Learning for Finance in Python
Calculate hypothetical portfolio
cash = 1000
algo_cash = [cash]
for r in test_returns:
cash *= 1 + r
algo_cash.append(cash)
# calculate performance for QQQ
cash = 1000 # reset cash amount
qqq_cash = [cash]
for r in returns_monthly['QQQ'].iloc[train_size:]:
cash *= 1 + r
qqq_cash.append(cash)
print('algo returns:', (algo_cash[-1] - algo_cash[0]) / algo_cash[0])
print('QQQ returns:', (qqq_cash[-1] - qqq_cash[0]) / qqq_cash[0])
algo returns: 0.5009443507049591
QQQ returns: 0.5186775933696601
DataCamp Machine Learning for Finance in Python
Plot the results
plt.plot(algo_cash, label='algo')
plt.plot(qqq_cash, label='QQQ')
plt.ylabel('$')
plt.legend() # show the legend
plt.show()
DataCamp Machine Learning for Finance in Python
MACHINE LEARNING FOR FINANCE IN PYTHON
Train your model!
DataCamp Machine Learning for Finance in Python
MACHINE LEARNING FOR FINANCE IN PYTHON
Final thoughts
Nathan George
Data Science Professor
DataCamp Machine Learning for Finance in Python
Toy examples
Tools for bigger data:
Python 3 multiprocessing
Dask
Spark
AWS or other cloud solutions
DataCamp Machine Learning for Finance in Python
Get more and better data
Data in this course:
From Quandl.com/EOD (free subset available)
Alternative and other data:
satellite images
sentiment analysis (e.g. PsychSignal)
analyst predictions
fundamentals data
DataCamp Machine Learning for Finance in Python
MACHINE LEARNING FOR FINANCE IN PYTHON
Be careful, and Godspeed!
