Logistic Regression¶
Titanic Survival¶
The sinking of the Titanic on April 15th, 1912 is one of the most tragic tragedies in history. The Titanic sank after colliding with an iceberg, killing 1502 out of 2224 passengers. The numbers of survivors were low due to the lack of lifeboats for all passengers and crew. Some passengers were more likely to survive than others, such as women, children, and upper-class. This case study analyzes what sorts of people were likely to survive this tragedy. The dataset includes the following:
- Pclass: Ticket class (1 = 1st, 2 = 2nd, 3 = 3rd)
- Sex: Sex
- Age: Age in years
- Sibsp: # of siblings / spouses aboard the Titanic
- Parch: # of parents / children aboard the Titanic
- Ticket: Ticket number
- Fare: Passenger fare
- Cabin: Cabin number
- Embarked: Port of Embarkation C = Cherbourg, Q = Queenstown, S = Southampton
- Target class: Survived: Survival (0 = No, 1 = Yes)
Importing the Relevant Libraries¶
In [1]:
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
sns.set()
Importing the Dataset¶
In [2]:
url = "https://datascienceschools.github.io/Machine_Learning/Classification_Models_CaseStudies/Train_Titanic.csv"
df = pd.read_csv(url)
df.head()
Out[2]:
Checking Missing Values¶
- Cabin & Embarked are unnecessary columns -> drop them after data visualisationIn [3]:
df.isnull().sum()
Out[3]:
Checking Missing Values (Heatmap)¶
In [4]:
sns.heatmap(df.isnull(),yticklabels=False,cbar=False,cmap='viridis')
plt.show()
Handling Missing Values (Age)¶
In [5]:
def Fill_Age(data):
age = data[0]
sex = data[1]
if pd.isnull(age):
if sex is 'male':
return 29
else:
return 27
else:
return age
df['Age'] = df[['Age','Sex']].apply(Fill_Age,axis=1)
df.isnull().sum()
Out[5]:
Explore the Dataset¶
Percentage of passengers Survived / Not Survived¶
In [6]:
survived = df[df['Survived'] == 1]
not_survived = df[df['Survived'] == 0]
print("Total =", len(df))
print("\nNumber of Survived passengers =", len(survived))
print("Percentage Survived = {:.2f}%".format(len(survived)*100/len(df)))
print("\nDid not Survive =", len(not_survived))
print("Percentage who did not survive = {:.2f}%".format(len(not_survived)*100/len(df)))
Installing cufflinks¶
In [7]:
!pip install cufflinks
Importing cufflinks¶
In [8]:
import cufflinks as cf
cf.go_offline()
Number of People Survived / Not Survived¶
In [9]:
survived = df[df['Survived']==1]['Survived'].value_counts()
dead = df[df['Survived']==0]['Survived'].value_counts()
df1 = pd.DataFrame([survived ,dead])
df1.index = ['Survived','Dead']
df1.iplot(kind='bar',barmode='stack', title='Number of Survived & Dead')
Number of People Survived based on Sex¶
- If you are a female,
- you have a higher chance of survivalIn [10]:
survived_sex = df[df['Survived']==1]['Sex'].value_counts()
dead_sex = df[df['Survived']==0]['Sex'].value_counts()
df1 = pd.DataFrame([survived_sex,dead_sex])
df1.index = ['Survived','Dead']
df1.iplot(kind='bar',barmode='stack', title='Survival by Sex')
Number of People Survived based on Class¶
- If you are a first class
- you have a higher chance of survivalIn [11]:
survived_pclass = df[df['Survived']==1]['Pclass'].value_counts()
dead_pclass = df[df['Survived']==0]['Pclass'].value_counts()
df1 = pd.DataFrame([survived_pclass, dead_pclass])
df1.index = ['Survived','Dead']
df1.iplot(kind='bar',barmode='stack', title='Survival by Pclass')
Number of People Survived based on Siblings Status¶
- If you have 1 sibling (SibSp = 1)
- you have a higher chance of survival compared to being alone (Parch = 0)In [12]:
survived_SibSp = df[df['Survived']==1]['SibSp'].value_counts()
dead_SibSp = df[df['Survived']==0]['SibSp'].value_counts()
df1 = pd.DataFrame([survived_SibSp, dead_SibSp])
df1.index = ['Survived','Dead']
df1.iplot(kind='bar',barmode='stack', title='Survival by Number of siblings / spouses aboard the Titanic')
Number of People Survived based on Parch Status (Parents/Children onboard)¶
- If you have 1 family member (Parch = 1)
- you have a higher chance of survival compared to being alone (Parch = 0)In [13]:
survived_Parch = df[df['Survived']==1]['Parch'].value_counts()
dead_Parch = df[df['Survived']==0]['Parch'].value_counts()
df1 = pd.DataFrame([survived_Parch, dead_Parch])
df1.index = ['Survived','Dead']
df1.iplot(kind='bar',barmode='stack', title='Survival by Number of parents / children aboard the Titanic')
Number of People Survived based on the port they emparked from¶
- Port of Embarkation C = Cherbourg, Q = Queenstown, S = Southampton
- If you embarked from port "C"
- you have a higher chance of survival compared to other ports!In [14]:
survived_Embarked = df[df['Survived']==1]['Embarked'].value_counts()
dead_Embarked = df[df['Survived']==0]['Embarked'].value_counts()
df1 = pd.DataFrame([survived_Embarked, dead_Embarked])
df1.index = ['Survived','Dead']
df1.iplot(kind='bar',barmode='stack', title='Survival by Port of Embarkation C = Cherbourg, Q = Queenstown, S = Southampton')
Number of People Survived based on Age¶
- If you are a baby
- you have a higher chance of survivalGrouping Age¶
In [15]:
df['Age_Group'] = pd.cut(df['Age'], bins=[0,5,10,20,30,40,50,60,70,80])
df.head()
Out[15]:
Survival by Age Group¶
In [16]:
survived_Age_Group = df[df['Survived']==1]['Age_Group'].value_counts()
dead_Age_Group = df[df['Survived']==0]['Age_Group'].value_counts()
df1 = pd.DataFrame([survived_Age_Group, dead_Age_Group])
df1.index = ['Survived','Dead']
df['Age'].iplot(kind='hist',bins=30, xTitle='Age',color='skyblue')
df['Age'].iplot(kind='box', xTitle='Age',color='lightgreen')
df1.iplot(kind='bar',barmode='stack', title='Survival by Age Group')
Number of People Survived based on Fare¶
- If you pay a higher fare
- you have a higher chance of survivalGrouping Fare¶
In [17]:
df['Fare_Group'] = pd.cut(df['Fare'], bins=[0, 50, 100, 200, 300, 600])
df.head()
Out[17]:
Survival by Fare Group¶
In [18]:
survived_Fare_Group = df[df['Survived']==1]['Fare_Group'].value_counts()
dead_Fare_Group = df[df['Survived']==0]['Fare_Group'].value_counts()
df1 = pd.DataFrame([survived_Fare_Group, dead_Fare_Group])
df1.index = ['Survived','Dead']
df['Fare'].iplot(kind='hist',bins=30, xTitle='Fare', color='lightgreen')
df['Fare'].iplot(kind='box', xTitle='Age',color='lightgreen')
df1.iplot(kind='bar',barmode='stack', title='Survival by Fare Group')
Handling Categorical Data - Dummy Variable(Sex)¶
- male: 1
- female: 0 In [19]:
df['Male'] = pd.get_dummies(df['Sex'], drop_first = True)
df.head()
Out[19]:
Drop Unnecessary Columns¶
In [20]:
df.drop(['PassengerId','Name', 'Sex','Ticket','Cabin', 'Embarked', 'Age_Group', 'Fare_Group' ], axis = 1 , inplace = True)
df.head()
Out[20]:
Rearrange Columns¶
In [21]:
df = df[['Pclass', 'Age', 'SibSp', 'Parch', 'Fare', 'Male', 'Survived']]
df.head()
Out[21]:
Declaring the Dependent & the Independent Variables¶
In [22]:
X = df.iloc[:,:-1].values
y = df.iloc[:,-1].values
Splitting the Dataset into the Training Set and Test Set¶
In [23]:
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, random_state = 11)
Feature Scaling¶
In [24]:
from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
X_train = sc.fit_transform(X_train)
X_test = sc.transform(X_test)
Training the Logistic Regression Model¶
In [25]:
from sklearn.linear_model import LogisticRegression
model = LogisticRegression(random_state = 0)
model.fit(X_train, y_train)
Out[25]:
Predicting the Test Set Results¶
In [26]:
y_pred = model.predict(X_test)
Confusion Matrix¶
In [27]:
from sklearn.metrics import confusion_matrix, accuracy_score
cm = confusion_matrix(y_test, y_pred)
accuracy = accuracy_score(y_test, y_pred)
print("Accuracy is: {:.2f}%".format(accuracy*100))
sns.heatmap(cm, annot = True, fmt="d")
plt.show()
Classification Report¶
In [28]:
from sklearn.metrics import classification_report
print(classification_report(y_test, y_pred))
k-Fold Cross Validation¶
In [29]:
from sklearn.model_selection import cross_val_score
accuracies = cross_val_score(estimator = model, X = X_train, y = y_train, cv = 10)
print("Accuracy: {:.2f} %".format(accuracies.mean()*100))
print("Standard Deviation: {:.2f} %".format(accuracies.std()*100))
Making Prediction (New Dataset)¶
Importing New Dataset¶
In [30]:
url = "https://datascienceschools.github.io/Machine_Learning/Classification_Models_CaseStudies/Test_Titanic.csv"
new_data = pd.read_csv(url)
new_data.head()
Out[30]:
Dropping unnecessary columns from New Dataset¶
In [31]:
new_data.drop(['PassengerId','Name', 'Ticket','Cabin', 'Embarked' ], axis = 1 , inplace = True)
new_data.head()
Out[31]:
Checking Missing Values¶
In [32]:
new_data.isnull().sum()
Out[32]:
Handling Missing Values ( Age & Fare)¶
In [33]:
def Fill_Age(data):
age = data[0]
sex = data[1]
if pd.isnull(age):
if sex is 'male':
return 29
else:
return 27
else:
return age
new_data['Age'] = new_data[['Age','Sex']].apply(Fill_Age,axis=1)
new_data = new_data.dropna(axis=0)
new_data.isnull().sum()
Out[33]:
Handling Categorical Data - Dummy Variable(Sex)¶
In [34]:
new_data['Male'] = pd.get_dummies(new_data['Sex'], drop_first = True)
new_data.drop(['Sex'], axis = 1, inplace = True)
new_data.head()
Out[34]:
Declaring Independent Variables¶
In [35]:
new_data_X = new_data.iloc[:,:].values
Feature Scaling (New Data)¶
In [36]:
new_data_X = sc.transform(new_data_X)
Predicting Dependent Variable (Survived)¶
In [37]:
new_data_y_pred = model.predict(new_data_X)
new_data['predicted_Survive'] = new_data_y_pred
new_data.head()
Out[37]:
Number of Predicted Survive / Not Servive¶
In [38]:
survive = new_data[new_data['predicted_Survive']==1]['predicted_Survive'].value_counts()
not_survive = new_data[new_data['predicted_Survive']==0]['predicted_Survive'].value_counts()
df1 = pd.DataFrame([survive , not_survive ])
df1.index = ['Survive','Not Survive']
df1.iplot(kind='bar',barmode='stack', title='Number of Predicted Survive & Not Survive')