regressionintro.py

# coding: utf-8

# In[20]:


import pandas as pd
import math, datetime
import numpy as np
from sklearn import preprocessing, cross_validation, svm
from sklearn.linear_model import LinearRegression
import matplotlib.pyplot as plt
from matplotlib import style
import pickle

style.use('ggplot')


# In[21]:


#Reading data
df = pd.read_csv('PETR4.SA.csv', header=0, index_col='Date', parse_dates=True)
print(df.head())


# In[22]:


# High - Low percentage
df['HL_PCT'] = (df['High'] - df['Low']) / df['Low'] * 100.0
# Percentage change
df['PCT_change'] = (df['Close'] - df['Open']) / df['Open'] * 100.0

df = df[['Close', 'HL_PCT', 'PCT_change', 'Volume']]


# In[23]:


# Select wich col will be used to forecast the prices
forecast_col = 'Close'
# Fill the NaN values with -99999, the algorithm will ignore these values
df.fillna(-99999, inplace=True)

# The number of predictions that will be done
forecast_out = int(math.ceil(0.01*len(df)))
print(forecast_out)

# Put all the values up the forecast_out number
df['label'] = df[forecast_col].shift(-forecast_out)


# In[24]:


X = np.array(df.drop(['label'], 1))
# Processing the entire data ignoring the 'label' col
X = preprocessing.scale(X)

# Array of predictions 
X_lately = X[-forecast_out:]
X = X[:-forecast_out]

df.dropna(inplace=True)
y = np.array(df['label'])


# In[25]:


# Setting up the machine learning algorithm
X_train, X_test, y_train, y_test = cross_validation.train_test_split(X, y, test_size=0.2)


# This is an example to use with big data, we don't need to train data every time we run the program
# with Pickle we can save the data trained before.
##clf = LinearRegression(n_jobs=-1)
##clf.fit(X_train, y_train)
##with open('linearregression.pickle', 'wb') as f:
##    pickle.dump(clf, f)

# And now we read the file created by pickle
pickle_in = open('linearregression.pickle', 'rb')
clf = pickle.load(pickle_in)

# This is an interesting value, this determines how much our algorithm can be "trusted" in percentage
accuracy = clf.score(X_test, y_test)
print(accuracy)


# In[26]:


# Predicting some values
forecast_set = clf.predict(X_lately)
print(forecast_set, accuracy, forecast_out)


# In[27]:



# Plotting a graph with real prices and forecasts generated by our algorithm 
df['Forecast'] = np.nan

last_date = df.iloc[-1].name
last_unix = last_date.timestamp()
one_day = 86400
next_unix = last_unix + one_day

for i in forecast_set:
    next_date = datetime.datetime.fromtimestamp(next_unix)
    next_unix += one_day
    df.loc[next_date] =  [np.nan for _ in range(len(df.columns)-1)] + [i]

df['Close'].plot()
df['Forecast'].plot()
plt.legend(loc=4)
plt.xlabel('Date')
plt.ylabel('Price')
plt.show()