#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Thu Sep 12 15:39:43 2019

@author: kylepaulus
"""


import monkdata as m
import drawtree_qt5 as u
import dtree as d
from operator import itemgetter
"""

Compute the average gain across all 6 attributes and print results

"""


for i in range(6):
    print('     ' + 'A' + str(i+1)+' =' + ' ' + str(d.averageGain(m.monk1, m.attributes[i])))


#print(d.averageGain(m.monk1, m.attributes[0]))
#print(d.averageGain(m.monk1, m.attributes[1]))
#print(d.averageGain(m.monk1, m.attributes[2]))
#print(d.averageGain(m.monk1, m.attributes[3]))
#print(d.averageGain(m.monk1, m.attributes[4]))
#print(d.averageGain(m.monk1, m.attributes[5]))


"""

Create a for loop for all 4 subsets of


"""

#monk1_A51 = d.select(m.monk1, m.attributes[4],1)
#monk1_A52 = d.select(m.monk1, m.attributes[4],2)
#monk1_A53 = d.select(m.monk1, m.attributes[4],3)
#monk1_A54 = d.select(m.monk1, m.attributes[4],4)

#import pandas as pd
#test = pd.DataFrame()
#for i in range(5):
#   for x in d.select(m.monk1, m.attributes[4],i):
#        test = test.append(pd.DataFrame(x.attribute, index=[0]))
#test = test.reset_index()
#
#
#
#a = d.select(m.monk1, m.attributes[4],i)

for i in range(5):
    for x in d.select(m.monk1, m.attributes[4],i):
        print('Monk1'+ ' '+ 'A5' + '=' + str(i) + str(x.attribute))

"""

Create a 4 loop that tests 6 attributes of A5 and computes average gain for
each attribute, then prints a label = to the value

Nested for loop 1-4, and a for loop 1-6


"""

#print(d.averageGain(monk1_A51, m.attributes[0]))
#print(d.averageGain(monk1_A51, m.attributes[1]))
#print(d.averageGain(monk1_A51, m.attributes[2]))
#print(d.averageGain(monk1_A51, m.attributes[3]))
#print(d.averageGain(monk1_A51, m.attributes[4]))
#print(d.averageGain(monk1_A51, m.attributes[5]))
#
#
#print(d.averageGain(monk1_A52, m.attributes[0]))
#print(d.averageGain(monk1_A52, m.attributes[1]))
#print(d.averageGain(monk1_A52, m.attributes[2]))
#print(d.averageGain(monk1_A52, m.attributes[3]))
#print(d.averageGain(monk1_A52, m.attributes[4]))
#print(d.averageGain(monk1_A52, m.attributes[5]))
#
#print(d.averageGain(monk1_A53, m.attributes[0]))
#print(d.averageGain(monk1_A53, m.attributes[1]))
#print(d.averageGain(monk1_A53, m.attributes[2]))
#print(d.averageGain(monk1_A53, m.attributes[3]))
#print(d.averageGain(monk1_A53, m.attributes[4]))
#print(d.averageGain(monk1_A53, m.attributes[5]))
#
#print(d.averageGain(monk1_A54, m.attributes[0]))
#print(d.averageGain(monk1_A54, m.attributes[1]))
#print(d.averageGain(monk1_A54, m.attributes[2]))
#print(d.averageGain(monk1_A54, m.attributes[3]))
#print(d.averageGain(monk1_A54, m.attributes[4]))
#print(d.averageGain(monk1_A54, m.attributes[5]))


#creat an empty list to call
monk_subsets = []
for i in range(4):
    monk_subsets.append(d.select(m.monk1, m.attributes[4],i+1))
#print(a5_list)


for i in range(4):
    print('Monk1  A5 Subsets = ' + str(i+1))
    for j in range(6):
        print('     ' + 'Monk1'+ ' ' + 'A'+ str(j+1) + ' ' + '=' +
              str(d.averageGain(monk_subsets[i], m.attributes[j])))

#for set in monk_subsets:
#	for attributes in m.attributes:
#		        print('     '+ str(d.averageGain(monk_subsets[i], m.attributes[i])))

"""

Assignment 5 Build and draw a tree and compute error


"""

#t = d.buildTree(m.monk3, m.attributes)
#error= 1-(d.check(t, m.monk3test))
#print(error)
#
#
#
#s = d.allPruned(t)[11]
#print(len(d.allPruned(s)))
#
#u.drawTree(s)






"""

Assignment 6 write a loop that prunes the trees and keeps pruning until the
model reaches a better value


"""

import random
def partition(data, fraction):
    ldata = list(data)
    random.shuffle(ldata)
    breakPoint = int(len(ldata) * fraction)
    return ldata[:breakPoint], ldata[breakPoint:]


monk1train, monk1val = partition(m.monk1, 0.6)

"""
while (all pruned function)error >= error(original tree)
    compute(error (all pruned))
print(tree)

"""
original_tree= d.buildTree(monk1train, m.attributes)
#pruned_tree = d.allPruned(original_tree)[0]
#error_original_tree = 1-(d.check(original_tree, monk1val))
#print(error_original_tree)
#error_pruned_tree = 1 - (d.check(pruned_tree, monk1val))
#print(error_pruned_tree)
#
#for i in range(100):
#    while error_original_tree <= error_pruned_tree:
#        pruned_tree = d.allPruned(original_tree)[i]
#        print(error_pruned_tree)



"""
Input: Tree and a validation set.
Returns a sorted list of tuples with pruned trees and accuracy like:
    [(tree, 0.8),(tree2, 0.85),...]
"""
def makelist(origtree, val):
    lst = d.allPruned(origtree)
    output = []
    for x in lst:
        output.append((x,d.check(x,val)))
    output = sorted(output,key=itemgetter(1))
    return output


print(makelist(original_tree, monk1val))

"""

Looks if the performance of a tree can be improved by pruning recursively.
Returns error number of the improved tree.


"""
def error_of_improved_tree(original_tree,val):
    error = 1 - d.check(original_tree, val)
    lst = makelist(original_tree, val)
    if d.check(original_tree, val) > lst[-1][1]:
        return error
    else:
        return error_of_improved_tree(lst[-1][0],val)

print(error_of_improved_tree(original_tree,monk1val))



def biggie(training, partition_parameter):
    lst = []
    for i in range(0,10):
        train, val = partition(training, partition_parameter)
        #original_tree = d.buildTree(train, m.attributes)
        lst.append(error_of_improved_tree)
    return lst

print(biggie(m.monk1, 0.3))