Machine learning - linear regression

import pandas as pd
import numpy as np


# y(x) = b + ax
def LinearFunction(theta, x):
    return theta[0] + theta[1] * x

#How well is our theta doing
def CalculateCost(theta, x, y):
    dataCount = len(y)
    total = 0
    for i in range(dataCount):
        total += ((LinearFunction(theta, x[i]) - y[i])**2)
    return (1.0 / (2 * dataCount)) * total

#Alpha - learning rate    Eps - when to stop
def GradientDescent(theta, x, y, alpha, eps):
    actualCost = CalculateCost(theta, x, y)
    dataCount = len(y)
    counter = 0
    while True:
        totalBeta = 0
        totalAlpha = 0

        #Count how much we need to change actual values
        for i in range(dataCount):
            totalBeta += (LinearFunction(theta, x[i]) - y[i])
            totalAlpha += (LinearFunction(theta, x[i]) - y[i]) * x[i]

        #change them acording to learning rate
        b = theta[0] - alpha/(float(dataCount)) * totalBeta
        a = theta[1] - alpha/(float(dataCount)) * totalAlpha

        #Update
        theta = [b,a]
        actualCost, previousCost = CalculateCost(theta, x, y), actualCost

        change = abs(previousCost - actualCost)
        #Show some info
        if counter % 100 == 0:
            print("Iteration: " + str(counter) + " Actual Cost: " + str(actualCost) + " Improved by: " + str(change))
        counter += 1
        #Check if we are done
        if change <= eps:
            print("Last: " + str(counter) + " Actual Cost: " + str(actualCost) + " Improved by: " + str(change))

            break
    return theta

#Scale
def Normalize(values):
    return (values - np.mean(values)) / np.std(values)

#Load data
trainData = pd.read_csv('train/train.tsv', sep='\t')
devDataInput = pd.read_csv('dev-0/in.tsv', sep='\t', header=None)[0]
testDataInput = pd.read_csv('test-A/in.tsv', sep='\t', header=None)[1]

#Prepare data
trainInput = Normalize(trainData['Powierzchnia w m2'])
trainOutput = trainData['cena']

devDataInput = Normalize(devDataInput)
print(len(devDataInput))
testDataInput = Normalize(testDataInput)
print(len(testDataInput))
#Solve and predict
solvedTheta = GradientDescent([4, -0.5], trainInput, trainOutput, 0.005, 0.0001)
print(solvedTheta)
predictedDev = LinearFunction(solvedTheta, devDataInput)
predictedTest = LinearFunction(solvedTheta, testDataInput)

np.savetxt('test-A/out.tsv', predictedTest, '%.0f')
np.savetxt('dev-0/out.tsv', predictedDev, '%.0f')