In [156]:

    

import pandas as pd
%matplotlib inline
from sklearn import datasets, tree, metrics
from sklearn.cross_validation import train_test_split
import matplotlib.pyplot as plt
import numpy as np

from sklearn.ensemble import RandomForestClassifier


    

In [157]:

    

import sklearn; print(sklearn.__version__)


    

    




0.17.1

In [158]:

    

iris = datasets.load_iris()


    

In [159]:

    

iris.keys()


    

    
Out[159]:





dict_keys(['feature_names', 'target', 'data', 'target_names', 'DESCR'])

In [160]:

    

X = iris.data[:,2:]
y = iris.target


    

In [161]:

    

X_train, X_test, y_train, y_test = train_test_split(X, y, stratify=y, random_state=42, test_size=0.25,train_size=0.75)
#What is random_state?
#What is stratify?


    

In [162]:

    

#What is this doing in the moon example exactly?
#X, y = make_moons(n_samples=100, noise=0.25, random_state=3)


    

In [163]:

    

forest = RandomForestClassifier(n_estimators=5, random_state=100)
forest.fit(X_train, y_train)


    

    
Out[163]:





RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',
            max_depth=None, max_features='auto', max_leaf_nodes=None,
            min_samples_leaf=1, min_samples_split=2,
            min_weight_fraction_leaf=0.0, n_estimators=5, n_jobs=1,
            oob_score=False, random_state=100, verbose=0, warm_start=False)

In [164]:

    

print("accuracy on training set: %f" % forest.score(X_train, y_train))
print("accuracy on test set: %f" % forest.score(X_test, y_test))


    

    




accuracy on training set: 0.990991
accuracy on test set: 0.948718

In [165]:

    

X_train, X_test, y_train, y_test = train_test_split(x,y,test_size=0.25,train_size=0.75)


    

In [166]:

    

dt = tree.DecisionTreeClassifier()


    

In [167]:

    

dt = dt.fit(X_train,y_train)


    

In [168]:

    

y_pred=dt.predict(X_test)


    

In [145]:

    

Accuracy_score = metrics.accuracy_score(y_test, y_pred)


    

In [146]:

    

Accuracy_score


    

    
Out[146]:





0.92105263157894735

In [ ]:

    

#Comments on RandomForestClassifiers & Original Decision Tree Model
#While the Random Trees result is consistent, varying depending how you choose the random_state or the n_estimators, 
#the result of the orgininal decision tree model varies a lot. 
#The random_state defines how random the versions of the data is that the modelling takes into consideration, and 
#the n_estimators regulates how many "random" datasets are used. It's fascinating to see how the this makes the 
#result so much more consistent than the orginal decision tree model.


    

In [147]:

    

#General commets on the homework
#I really enjoyed this homework and it really helped me understand, what is going on under the hood. 
#I found this reading while I was doing the homework. It looks nice to go deeper? Do you know the
#guy? https://github.com/amueller/introduction_to_ml_with_python
#I feel I now need practice on real life dirty data sets, to fully understand how predictions models 
#can work. I take my comments back, that I can't see how I can implement this into my reporting. I can. But how 
#can I do this technically? i.e. with the data on PERM visas? Say input nationality, wage, lawyer, job title, and get a reply what the chances could be of 
#getting a work visa? I also feel a little shaky on how I need to prep my data to feed in it into the predictor 
#correctly.


    

In [ ]:

    

#Comments on classifier
#Questions:
#Not sure why it's 10fold cross validation, cv is set at 5?
#Why are we predicting the


    

In [ ]: