Module 9 - Tutorial - Decision Trees

import pandas as pd

import numpy as np

import matplotlib.pyplot as plt

import seaborn as sns

%matplotlib inline

from sklearn import tree

from sklearn.model_selection import train_test_split

from sklearn.metrics import confusion_matrix

from sklearn.metrics import classification_report

location = titanic.xls

df = pd.read_excel(location)

df.head()

#find columns that have missing values

df.isnull().sum()

#fill missing values for age based on survival status, sex, and passenger class

df['age'].fillna(df.groupby(['survived', 'sex', 'pclass'])['age'].transform('mean'), inplace=True)

#only 2 missing values so we'll fill with most common embarkation point

df['embarked'].value_counts()

#fill missing values

df['embarked'].fillna('S', inplace=True)

df.isnull().sum()

modeldf = df.drop(['name','ticket','fare', 'cabin', 'boat', 'body', 'home.dest'], axis=1)

#columns left in our dataframe

modeldf.columns

#dummy variables for passenger class embarkation port

#get_dummies will auto-drop columns that dummies were created from

modeldf = pd.get_dummies(data=modeldf, columns=['pclass','embarked'])

#modeldf.head()

#change sex values to binary

#female=0, male=1

modeldf['sex'] = modeldf['sex'].map({'female':0, 'male':1})

#modeldf.head()

#create new column based on number of family members

#drop sibsp and parch columns

modeldf['family_num'] = modeldf['sibsp'] + modeldf['parch']

modeldf.drop(['sibsp', 'parch'], axis=1, inplace=True)

#modeldf.head()

modeldf['TravelAlone']=np.where((modeldf['family_num'] > 0), 0, 1)

#modeldf.head()

#extract target variable

#make copy of 'survived' column

y = modeldf['survived']

#copy of modeldf without 'survived' column

X = modeldf.drop(['survived'], axis=1)

#80% for training data, 20% for test data

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=15)

#assign decision tree function to model variable

tree = tree.DecisionTreeClassifier()

#develop model using training data

#defining arguments in the model can help prevent overfitting

tree.fit(X_train, y_train)

#accuracy score of model on training data

tree.score(X_train, y_train)

#run the predictions on the test data

y_predict = tree.predict(X_test)

#accuracy score of model on test data

tree.score(X_test, y_test)

#look at true and false predictions

pd.DataFrame(

    confusion_matrix(y_test, y_predict),

    columns=['Predicted Not Survival', 'Predicted Survival'],

    index=['True Not Survival', 'True Survival']

)

#from precision column, model is better at predicting passengers that do not survive

print(classification_report(y_test, y_predict))