from statistics import mean
import numpy as np
import matplotlib.pyplot as plt
from matplotlib import style as stl


#xf=np.array([1,2,3,4,5,6],dtype=np.float64)
#yf=np.array([5,4,6,5,6,7],dtype=np.float64)


stl.use("fivethirtyeight")
xf=np.array([36961,43621,15694,36231,29945,40588,75255,37709,30899,25486,37497,40398,74105,76725,18317],dtype=np.float64)
yf=np.array([2503,2992,1042,2487,2014,2805,5062,2643,2126,1784,2641,2766,5047,5312,1215],dtype=np.float64)



xe= np.array([35680,42514,15162,35298,29800,40255,74532,37464,31030,24843,36172,39552,72545,75352,18031],dtype=np.float64)
ye=np.array([2217,2761,990,2274,1865,2606,4805,2396,1993,1627,2375,2560,4597,4871,1119],dtype=np.float64)

def standard_error(x, y, m, b):
    toterror =0
    for i in range(len(x)):
        toterror+=(y[i]-(x[i]*m+b))**2
    toterror=toterror/len(x)
    return toterror

def gradient(x, y, m, b):
    dm =0
    db=0
    N=float(len(x))
    for i in range(0,len(x)):
        print("gradient.... dm= ",dm,"db==",db)
        dm+=-(2/N)*(x[i]*(y[i]-(x[i]*m+b)))
        db+=-(2/N)*(y[i]-m*x[i]+b)

       # db= -(2/N) * (y[i] - ((m*x[i]) + b))
        #dm= -(2/N) * x[i] * (y[i] - ((m * x[i]) + b))
    return dm,db


def gradient_decent(x, y,error_threshold):
    m=0
    b=0
    error= 100
    last_error=error+1
    while(error>error_threshold ):
        print("Error: "+str(error))
        dm,db = gradient(x,y,m,b)
        print("dm= ",dm,"db==",db)
        error = standard_error(x,y,m,b)
        m=m-dm*0.2
        b=b-db*0.2
        #b=b+db
        last_error =error
        error = standard_error(x,y,m,b)
    return m,b






def line_fit(x,y):
    m=(( (mean(x) * mean(y) ) - mean(x*y) ) /

       ( (mean(x)*mean(x)) - mean(x*x) ) )
    b= mean(y)- m*mean(x)
    return m,b


def square_error(y_orig, y_line):

    return sum( (y_line-y_orig)**2)


def R2(y_orig,y_line):

    y_mean_line = [mean(y_orig) for y in y_orig]
    SE_reg = square_error(y_orig, y_line)
    SE_mean = square_error(y_orig, y_mean_line)

    return 1-(SE_reg/SE_mean)



def printus_maximus(x,y):
    m,b= line_fit(x,y)
    regression_line=[(m*x)+b for x in x]
    r_squared = R2(y, regression_line)
    print("R^2 = ",x, r_squared  )
    print("Result = ",m,b)
    plt.scatter(x,y)
    plt.plot(x,regression_line)


def printus_maximus2dot0(x,y):
    m,b= gradient_decent(x,y,0.0005)
    regression_line=[(m*x)+b for x in x]
    plt.scatter(x,y)
    plt.plot(x,regression_line)



printus_maximus2dot0(xf,yf)