import matplotlib.pyplot as plt
import numpy as np

# my dataframe is df

# plot for see if there is a linear relation
plt.scatter(col1, col2)
plt.show()

# create training and test data set
rand = np.random.rand(len(df)) < 0.8
train = cdf[rand]
test = cdf[~rand]

# MODEL
from sklearn import linear_model
regr = linear_model.LinearRegression()
train_x = np.asanyarray(train[['col1']])
train_y = np.asanyarray(train[['col2']])
regr.fit (train_x, train_y)

# coefficients
print ('Coefficients: ', regr.coef_)
print ('Intercept: ',regr.intercept_)

# plot output
plt.scatter(train.col1, train.col2)
plt.plot(train_x, regr.coef_[0][0]*train_x + regr.intercept_[0], color='red')

# EVALUATION
from sklearn.metrics import r2_score

test_x = np.asanyarray(test[['col1']])
test_y = np.asanyarray(test[['col2']])
test_y_ = regr.predict(test_x)

print("Mean absolute error: %.2f" % np.mean(np.absolute(test_y_ - test_y)))
print("Residual sum of squares (MSE): %.2f" % np.mean((test_y_ - test_y) ** 2))
print("R2-score: %.2f" % r2_score(test_y_ , test_y) )