huber_reg.py

#import library
import pandas as pd
import numpy as np
import sys
from sklearn.preprocessing import Imputer
from sklearn.model_selection import train_test_split
from sklearn.linear_model import ElasticNet
from sklearn.linear_model import HuberRegressor
from matplotlib import pyplot
import crayons

#loading datasets
data = pd.read_csv(sys.argv[1])
data = pd.DataFrame(data)

# Finding the right column for the trading
df = data[['Prev Close','Last Price']]
y = data['Close Price']
pc = list(data['Prev Close'])
X = pd.DataFrame(df)
#splitting the datasets into train and test set
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.15, shuffle=False)

#calling model
clf = HuberRegressor(alpha = 0.001)
fit_model = clf.fit(X_train , y_train)

#predicting on test data
y_pred = fit_model.predict(X_test)
y_pred = pd.Series(y_pred)
num = len(list(y_pred))
print(y_pred[num-1])
y_test1 = y_test.shift(1) #shifting data one step into future
y_true = y_test-y_test1

#converting negative data into 0 and postive data into 1.
y_true[y_true<0] = 0 # '0' stands for down
y_true[y_true>0] = 1 # '1' stands for up

y_true  = pd.DataFrame(y_true)

#replacing the NAN value with 0.   
y_true = y_true.fillna(round(np.mean(y_true)))
#same applying for predicted value
y_pred1 = y_pred.shift(1) #shifting the data one step into the future

y_fake = y_pred-y_pred1

#converting negative data into 0 and postive data into 1.
y_fake[y_fake<0] = 0
y_fake[y_fake>0] = 1

#changing data into DataFrame
y_fake  = pd.DataFrame(y_fake)
#replacing the NAN value with 0.   
y_fake = y_fake.fillna(0)
#changin the data into a numpy array
y_true = np.array(y_true)
y_fake = np.array(y_fake)

#reshaping the data into 1 dimensional array
y_true = np.reshape(y_true, len(y_true))
y_fake = np.reshape(y_fake, len(y_fake))

#comparing corrosponding item in the given two array
correct_pred = sum(len(set(i)) == 1 for i in zip(y_true, y_fake))

#calculating the accuracy of the model on the basis of up and down
accuracy = (correct_pred/len(y_true))*100

#printing the results
print(crayons.yellow(f'\t[*] Total number of forward forecasting  => {len(y_true)}\n', bold=True))
print(crayons.yellow(f'\t[*] Showing Actual and Predicted value in terms of 1s and 0s\n', bold=True))
print(crayons.yellow(f'\t[*] Actual value of Close Price => {y_true}\n', bold=True))
print(crayons.red(f'\t[*] Predicted vaue of Close Price => {y_fake}\n', bold=True))
print(crayons.blue(f'\t[*] ACCURACY : => {accuracy}', bold=True))
# change this it into either up and down base on the if the predictionis is negative or positive


predict_trend = (y_pred[num-1] - pc[-1:]).round()

if predict_trend >= 0:
    print(crayons.yellow(f'\t[*] The forecast trend is UP\n', bold=True))
else:
    print(crayons.yellow(f'\t[*] The forecast trend is DOWN\n', bold=True))

# Plot
actual_line = pyplot.plot(y_true, marker='s', label='Actual Price')
predicted_line = pyplot.plot(y_fake, color='red', marker='o', label='Predicted Price')
pyplot.legend(loc='upper left')
pyplot.ylabel('Trend1=up and 0=down]')
pyplot.xlabel('Data')
pyplot.title('Forecast Results')
pyplot.grid()
pyplot.show()