import pylab as pl
import numpy as np
from sklearn import svm, datasets

print("\nImporting training data...")
#removing pybrain dataset items
#bring in data from training file
f = open("zipcode.tra")
inp = []
tar = []
i=0
#for line in f.readlines():
#    inp.append(list(map(float, line.split()))[0:16])
#    tar.append(int(list(map(float, line.split()))[16])-1)
for line in f.readlines():
    if i == 0:
        #reset lists
        sample = []
        target = []
    if i < 3:
        sample.extend(list(map(float, line.split()))[0:5])
    if i == 3:
        sample.extend(list(map(float, line.split()))[0:1])
        tar.append(int(list(map(float, line.split()))[1])-1)
        inp.append(sample)
        i = 0
    else:
        i += 1    
 
print("Training rows: %d " %  len(inp) )
print("Input dimensions: ", np.shape(inp), ", Output dimensions: ", np.shape(tar))
#raw_input("Press Enter to view training data...")
#print(traindata)
print("\nFirst sample: ", inp[0], tar[0])

#sk-learn SVM
svm_model = svm.SVC()#kernel='linear')
svm_model.fit(inp,tar)

#show predictions
#out = net.activateOnDataset(ds)
out = svm_model.predict(inp)
#print(test)
class_predict = [0.0 for i in range(len(out))]
predict_correct = []
pred_corr_class0 = []
pred_corr_class1 = []

#NOTE - there are 10 classes in this one!
for index, row in enumerate(out):
    class_predict[index] = row
for N in range(len(class_predict)):
    if class_predict[N] == tar[N]:
        predict_correct.append(1)
    else:
        predict_correct.append(0)

print ("SVM Kernel: ", svm_model.kernel)
print ("\nCorrectly classified: %d, Percent: %f, Error: %f" % (sum(predict_correct),sum(predict_correct)/float(len(predict_correct)),1-sum(predict_correct)/float(len(predict_correct))))
#print ("Class 0 correct: %d, Percent: %f, Error: %f" % (sum(pred_corr_class0),sum(pred_corr_class0)/float(len(pred_corr_class0)),1-sum(pred_corr_class0)/float(len(pred_corr_class0))))
#print ("Class 1 correct: %d, Percent: %f, Error: %f" % (sum(pred_corr_class1),sum(pred_corr_class1)/float(len(pred_corr_class1)),1-sum(pred_corr_class1)/float(len(pred_corr_class1))))



raw_input("\nPress Enter to start testing...")

print("\nImporting testing data...")
#bring in data from testing file
f = open("zipcode.tst")
inp_tst = []
tar_tst = []
i=0
for line in f.readlines():
    if i == 0:
        #reset lists
        sample = []
        target = []
    if i < 3:
        sample.extend(list(map(float, line.split()))[0:5])
    if i == 3:
        sample.extend(list(map(float, line.split()))[0:1])
        tar_tst.append(int(list(map(float, line.split()))[1])-1)
        inp_tst.append(sample)
        i = 0
    else:
        i += 1    
    
 
print("Testing rows: %d " %  len(inp_tst) )
print("Input dimensions: ", np.shape(inp_tst), ", Output dimensions: ", np.shape(tar_tst))
print("\nFirst sample: ", inp_tst[0], tar_tst[0])

print("\nTesting...")

#show predictions
#out = net.activateOnDataset(ds)
out_tst = svm_model.predict(inp_tst)
#print(test)
class_predict = [0.0 for i in range(len(out_tst))]
predict_correct = []
pred_corr_class0 = []
pred_corr_class1 = []
for index, row in enumerate(out_tst):
    class_predict[index] = row
for N in range(len(class_predict)):
    if class_predict[N] == tar_tst[N]:
        predict_correct.append(1)
    else:
        predict_correct.append(0)

print ("\nTest Data Correctly classified: %d, Percent: %f, Error: %f" % (sum(predict_correct),sum(predict_correct)/float(len(predict_correct)),1-sum(predict_correct)/float(len(predict_correct))))
#print ("Class 0 correct: %d, Percent: %f, Error: %f" % (sum(pred_corr_class0),sum(pred_corr_class0)/float(len(pred_corr_class0)),1-sum(pred_corr_class2)/float(len(pred_corr_class0))))
#print ("Class 1 correct: %d, Percent: %f, Error: %f" % (sum(pred_corr_class1),sum(pred_corr_class1)/float(len(pred_corr_class1)),1-sum(pred_corr_class1)/float(len(pred_corr_class1))))