import numpy as np

def d(test,m):
    epred = []
    mpred = []
    for i in range(len(test)):
        x = []
        euc = []
        man = []
        for f in range(len(test[0])-1):
            x.append(test[i,f+1])
        for c in range(3):
            me = m[:,c]
            euc.append(np.sqrt(sum(([(i - j)**2 for i, j in zip(x, me)]))))
            man.append(abs(sum(x)-sum(m[:,c])))
        epred.append(euc.index(min(euc)))
        mpred.append(man.index(min(man)))
        #print("predict by euclidean - "+str(epred[i]))
        #print("predict by manhattan - "+str(mpred[i]))
        #print("real - "+str(test[i,0]))

    re = 0
    rm = 0
    for i in range(len(test)):
        if epred[i]==test[i,0]:
            re+=1
        if mpred[i]==test[i,0]:
            rm+=1
    print("Euclidian distance accuracy - "+str(round(100*re/len(test)))+"%")
    print("Manhattan distance accuracy - "+str(round(100*rm/len(test)))+"%")


file = open("fisher.txt", "r")
lines = []
for i,line in enumerate(file):
    if i!=0:
        lines.append([int(x) for x in line.strip().split("\t")])

data = np.array(lines)
train = data[0:119,:]
test = data[120:,:]

T0 = data[data[:,0] == 0,1:]
T1 = data[data[:,0] == 1,1:]
T2 = data[data[:,0] == 2,1:]

m = []
for i in range(4):
    m.append([np.mean(T0[:,i]),np.mean(T1[:,i]),np.mean(T2[:,i])])
m = np.array(m)

X1 = train[:,1]
X2 = train[:,2]
X3 = train[:,3]
X4 = train[:,4]

Mu1 = np.mean(X1)
Mu2 = np.mean(X2)
Mu3 = np.mean(X3)
Mu4 = np.mean(X4)

d(test,m)


 #PCA

X1p = X1-Mu1
X2p = X2-Mu2
X3p = X3-Mu3
X4p = X4-Mu4

cm = np.cov([X1p,X2p,X3p,X4p])

w, v = np.linalg.eig(cm)
index = np.argsort(-w)
fv = v[:,index]
rfv = np.transpose(fv)
rzm = np.array([X1p, X2p, X3p, X4p])
fd = np.transpose(np.matmul(rfv, rzm))

f1 =fd[:,0]
f2 =fd[:,1]
f3 =fd[:,2]

for i in range(len(test)):
    for f in range(3):
        test[i,f+1]=fd[i,f]
    np.delete(test[i],4)

d(test,m)