#Data Mining Python Examples
#
#Scikit-learn
#
#Scikit-Learn tutorial:
#http://scikit-learn.org/stable/tutorial/basic/tutorial.html

#####################################
#Importing Modules sklearn and numpy#
#####################################

from sklearn import *
import numpy as np

#arange(5)
numpy.arange(5)
np.arange(5)


#################################
#Built in Data Sets from sklearn#
#################################

#Importing iris data
iris=datasets.load_iris()
iris
type(iris)

#Iris independent variables
iris.data
type(iris.data)
np.shape(iris.data)
iris.data.shape

#Iris Dependent variable (Target Variable)
iris.target
type(iris.target)
iris.target.shape


###########################
#Training and Testing Data#
###########################

iris.data.shape
iris.data.shape[0]
np.arange(iris.data.shape[0])

train,test = cross_validation.train_test_split(np.arange(iris.data.shape[0]),
                                                test_size=0.3,
                                                random_state=5366)
                                                
train
test                                                

iris.data[train,:]
iris.data[test,:]

iris.target[train]
iris.target[test]


                                              
################                                                
#Decision Trees#
################

model=tree.DecisionTreeClassifier()
model=model.fit(iris.data[train,:],iris.target[train])

predspecies=model.predict(iris.data[test,:])

metrics.confusion_matrix(iris.target[test],predspecies)
metrics.accuracy_score(iris.target[test],predspecies)

                              
#Plotting the tree (Requires pydot and graphviz)
#http://www.graphviz.org/
                              
from sklearn.externals.six import StringIO
import pydot  
dot_data = StringIO() 
tree.export_graphviz(model, out_file=dot_data) 
graph = pydot.graph_from_dot_data(dot_data.getvalue()) 
graph.write_pdf("IrisTree.pdf")


#########################
#K-Fold Cross-validation#
#########################

scores=cross_validation.cross_val_score(
            model,iris.data,iris.target,cv=10)
scores
np.mean(scores)


#####################
#K-Nearest Neighbors#
#####################

model=neighbors.KNeighborsClassifier()
model=model.fit(iris.data[train,:],iris.target[train])

predspecies=model.predict(iris.data[test,:])

metrics.confusion_matrix(iris.target[test],predspecies)
metrics.accuracy_score(iris.target[test],predspecies)


######################
#Gaussian Naive Bayes#
######################

#Note that we are using Gaussian Naive Bayes since the
#features in this problem (Sepal Length, Sepal Width, etc)
#are quantitative.  

#See the documentation for examples with multinomial Naive Bayes, etc.

model=naive_bayes.GaussianNB()
model=model.fit(iris.data[train,:],iris.target[train])

predspecies=model.predict(iris.data[test,:])

metrics.confusion_matrix(iris.target[test],predspecies)
metrics.accuracy_score(iris.target[test],predspecies)


########################
#Support Vector Machine#
########################

model=svm.SVC()
model=model.fit(iris.data[train,:],iris.target[train])

predspecies=model.predict(iris.data[test,:])

metrics.confusion_matrix(iris.target[test],predspecies)
metrics.accuracy_score(iris.target[test],predspecies)


#########
#Bagging#
#########


model=ensemble.BaggingClassifier()
model=model.fit(iris.data[train,:],iris.target[train])

predspecies=model.predict(iris.data[test,:])

metrics.confusion_matrix(iris.target[test],predspecies)
metrics.accuracy_score(iris.target[test],predspecies)


##########
#Boosting#
##########


model=ensemble.AdaBoostClassifier()
model=model.fit(iris.data[train,:],iris.target[train])

predspecies=model.predict(iris.data[test,:])

metrics.confusion_matrix(iris.target[test],predspecies)
metrics.accuracy_score(iris.target[test],predspecies)


###############
#Random Forest#
###############


model=ensemble.RandomForestClassifier()
model=model.fit(iris.data[train,:],iris.target[train])

predspecies=model.predict(iris.data[test,:])

metrics.confusion_matrix(iris.target[test],predspecies)
metrics.accuracy_score(iris.target[test],predspecies)