Assignment 1

Use the pseudocode you came up with in class to write your own 5-fold cross-validation function that splits the data set into 5 equal-sized sets
Don't forget to shuffle the input before assigning to sets
You can use the fit(), predict(), and score() functions of your model in your functions
Test the results with the sklearn cross_val_score
In your PR, discuss what challenges you had creating this function and if it helped you better understand cross validation

Pseudocode

function for 5-fold cross validation:

    Shuffle the index of the data. 

    Split the data in 5 parts.

    Loop, repeated 5 times:

        Train the system using four parts. 

        Compute the accuracy using the not used part.

        Store it in a dictionary.

    Compute the average for the 5 accuracy values from the loop.



In [50]:

    
import pandas as pd
%matplotlib inline
from sklearn import datasets
from sklearn import tree
import matplotlib.pyplot as plt
from random import shuffle



In [157]:

    
def lets_shake(x,y):
    zipped = zip(x,y) # Let's bind the x and the y together before shakin'.
    zipped_list = list(zipped) 
    shuffle(zipped_list)
    x, y = zip(*zipped_list)
    return x, y



In [222]:

    
def the_splitter(x,y,folds):
    divider = int(len(x)/folds)
    group_dict = {}
    
    for i in range(0,folds):
        temp_list = list()
        for j in range(i*divider,(i+1)*divider):
            temp_list.append((x[j], y[j]))
        group_dict['group' + str(i)] = temp_list
    return group_dict
# This function still neglect a small rest of the values.



In [246]:

    
def the_test(groups,folds):
    dt = tree.DecisionTreeClassifier()
    for group in groups:
        dt = dt.fit(x,y)
        #dt.predict()
        #score()



In [231]:

    
x = [[1, 1], [2, 2], [3, 3], [4, 4], [5,5], [6,6], [7,7], [8,8], [9,9], [10,10]]
y = [0,2,1,1,0,1,2,1,1,0]



In [232]:

    
x,y = lets_shake(x,y)
groups = the_splitter(x,y,5)



In [236]:

    
groups









    Out[236]:





{'group0': [([3, 3], 1), ([4, 4], 1)],
 'group1': [([7, 7], 2), ([5, 5], 0)],
 'group2': [([9, 9], 1), ([10, 10], 0)],
 'group3': [([8, 8], 1), ([1, 1], 0)],
 'group4': [([2, 2], 2), ([6, 6], 1)]}



In [ ]:



In [ ]:



In [ ]: