Skip to content

Instantly share code, notes, and snippets.

@greencoder
Last active July 27, 2022 19:32
Show Gist options
  • Star 3 You must be signed in to star a gist
  • Fork 4 You must be signed in to fork a gist
  • Save greencoder/ab37304b6d47e6d1e55b4adf96ea7b47 to your computer and use it in GitHub Desktop.
Save greencoder/ab37304b6d47e6d1e55b4adf96ea7b47 to your computer and use it in GitHub Desktop.
SKLearn Linear Regression Stock Price Prediction
from __future__ import print_function
import numpy as np
import pandas as pd
import talib as ta
import pandas_datareader as web
from sklearn.linear_model import LinearRegression
from sklearn.linear_model import Ridge
from sklearn.model_selection import cross_val_score
from sklearn.model_selection import train_test_split
from sklearn.naive_bayes import GaussianNB
from sklearn.pipeline import make_pipeline
from sklearn.preprocessing import StandardScaler
from sklearn.tree import DecisionTreeRegressor
def make_prediction(quotes_df, estimator):
# Make a copy of the dataframe so we don't modify the original
df = quotes_df.copy()
# Add the five day moving average technical indicator
df['MA_5'] = ta.MA(df['Close'].values, timeperiod=5, matype=0)
# Add the twenty day moving average technical indicator
df['MA_20'] = ta.MA(df['Close'].values, timeperiod=20, matype=0)
# Add the fifty day moving average technical indicator
df['MA_50'] = ta.MA(df['Close'].values, timeperiod=50, matype=0)
# Add the Bollinger Bands technical indicators
df['BOL_Upp'], df['BOL_Mid'], df['BOL_Low'] = ta.BBANDS(df['Close'].values,
timeperiod=20, nbdevup=2, nbdevdn=2, matype=0)
# Add the Relative strength index technical indicator
df['RSI'] = ta.RSI(df['Close'].values, 14)
# Add the Simple Moving Average (Fast & Slow) technical indicators
df['SMA_Fast'] = ta.SMA(df['Close'].values, 5)
df['SMA_Slow'] = ta.SMA(df['Close'].values, 20)
# Add the percent change of the daily closing price
df['ClosingPctChange'] = df['Close'].pct_change()
# Get today's record (the last record) so we can predict it later. Do this
# before we add the 'NextDayPrice' column so we don't have to drop it later
df_today = df.iloc[-1:, :].copy()
# Create a column of the next day's closing prices so we can train on it
# and then eventually predict the value
df['NextClose'] = df['Close'].shift(-1)
# Get rid of the rows that have NaNs
df.dropna(inplace=True)
# Decide which features to use for our regression. This will allow us to
# tweak things during testing
features_to_fit = ['Open', 'High', 'Low', 'Close', 'Volume', 'MA_20', 'MA_50',
'RSI', 'SMA_Fast', 'SMA_Slow', 'BOL_Upp', 'BOL_Mid', 'BOL_Low', 'ClosingPctChange']
# Create our target and labels
X = df[features_to_fit]
y = df['NextClose']
# Create training and testing data sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.4,
random_state=42)
# Do ten-fold cross-validation and compute our average accuracy
cv = cross_val_score(estimator, X_test, y_test, cv=10)
print('Accuracy:', cv.mean())
# Fit the regressor with the full dataset to be used with predictions
estimator.fit(X, y)
# Predict today's closing price
X_new = df_today[features_to_fit]
next_price_prediction = estimator.predict(X_new)
# Return the predicted closing price
return next_price_prediction
# Choose which company to predict
symbol = 'AAPL'
# Import a year's OHLCV data from Google using DataReader
quotes_df = web.data.DataReader(symbol, 'google')
# Predict the last day's closing price using linear regression
print('Unscaled Linear Regression:')
linreg = LinearRegression()
print('Predicted Closing Price: %.2f\n' % make_prediction(quotes_df, linreg))
# Predict the last day's closing price using Linear regression with scaled features
print('Scaled Linear Regression:')
pipe = make_pipeline(StandardScaler(), LinearRegression())
print('Predicted Closing Price: %.2f\n' % make_prediction(quotes_df, pipe))
# Predict the last day's closing price using ridge regression
print('Unscaled Ridge Regression:')
ridge = Ridge()
print('Predicted Closing Price: %.2f\n' % make_prediction(quotes_df, ridge))
# Predict the last day's closing price using ridge regression and scaled features
print('Scaled Linear Regression:')
ridge_pipe = make_pipeline(StandardScaler(), Ridge())
print('Predicted Closing Price: %.2f\n' % make_prediction(quotes_df, ridge_pipe))
# Predict the last day's closing price using decision tree regression
print('Unscaled Decision Tree Regressor:')
tree = DecisionTreeRegressor()
print('Predicted Closing Price: %.2f\n' % make_prediction(quotes_df, tree))
# Predict the last day's closing price using Gaussian Naive Bayes
print('Unscaled Gaussian Naive Bayes:')
nb = GaussianNB()
print('Predicted Closing Price: %.2f\n' % make_prediction(quotes_df, nb))
@greencoder
Copy link
Author

If anyone comes across this gist and is interested in trading notes and implementation ideas, please let me know. I'm getting high accuracy on my test set, but real-world performance is not great.

@jeffreyscottgraham
Copy link

The Gaussian Naive Bayes predictor is defined on line 117, but is not passed to 'make_prediction' on line 118.
Correcting that results in a compile error 'ValueError: Unknown label type....'
Can you fix this?

@rmihir96
Copy link

rmihir96 commented Mar 6, 2018

I want to implement a project to find (predict) points scored by each player for the coming gameweek. I am targeting Fantasy Premier League(FPL) for this. What I exactly want to do is using previous data based on some features the model should predict the total points to be scored by a player. Example : Consider a player - Lukaku. Suppose for Lukaku, I have data for 28 game weeks and I want to find out how many points would Lukaku score for 29th game week(consider factors like goals scored, assists, minutes played, etc).

I had thought about implementing Linear Regression for which I can train the model for 28 game weeks but then I dont know how or what should be my input for the 29th week so as to predict points for that week.

Can you help me with this?

@grx7
Copy link

grx7 commented May 3, 2018

I got Bayesian technical paper but needs to be coded.. Wanna collab on it? :)

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment