btc-prices-5.py

import numpy as np
import csv
import matplotlib.pyplot as plt

# 1. Extract Bitcoin prices and number of Google searches.
bitcoin_interest = {}

with open('bitcoin-interest.csv') as f:
    reader = csv.reader(f)
    for row in reader:
        bitcoin_interest[row[0]] = row[1]
        
price_and_interest = []
with open('market-price.csv') as f:
    reader = csv.reader(f)
    for row in reader:
        # 2. Match the price with search date.
        if bitcoin_interest.get(row[0], None):
            price_and_interest.append([int(bitcoin_interest.get(row[0])),  float(row[1])])
           def calculate_error(points, m, b):
    # Error is calculated by the average distance from the points to the line
    error = 0
    for i in range(0, len(points)):
        x = points[i, 0]
        y = points[i, 1]
        
        # Moving y to the other side of the equation
        # y = mx + b ->  = mx + b - y
        error += (y - (m*x + b))**2
    
    # Calculating average
    return error / float(len(points))

def gradient_descent(points, b, m, learning_rate):
    m_gradient = 0
    b_gradient = 0
    
    N = float(len(points))
    for i in range(0, len(points)):
        x = points[i, 0]
        y = points[i, 1]
        
        b_gradient += -(2/N) * (y - ((m * x) + b))
        m_gradient += -(2/N) * x * (y - ((m * x) + b))

    m_updated = m - (m_gradient * learning_rate)
    b_updated = b - (b_gradient * learning_rate)
    
    return b_updated, m_updated


# 3.Set our hyper paremeters: epoch, learning rate, m and b.

learning_rate = 0.0001
epochs = 1000
start_m = 0
start_b = 0

for i in range(epochs):
    # 4. Calculate the error so we can see our algorirthm learning.
    error = calculate_error(data, start_m, start_b)
    if (i % 100):
        print("Error rate: " + str(error))
    # 5. Calculate the gradient:
    start_m, start_b = gradient_descent(data, start_m, start_b, learning_rate)
    
print("Final m: " + str(start_m))
print("Final b: " + str(start_b))