TITANIC SURVIVAL CLASSIFICATION

TASK:

To predict whether a person will survive the titanic crash given a set of features.

##DATASET:

• The training dataset can be downloaded here.
• The testing dataset can be downloaded here

FEATURES:

FEATURES	DESCRIPTION
* SEX	Male or female
* AGE	Age of the person
* N_SIBLINGS_SPOUSES	Siblings and spouses on board
* PARCH	Parents on board
* FARE	Price of the ticket to board the ship
* CLASS	First or Second or Third class
* DECK	The location of his/her cabin
* EMBARK_TOWN	The town from which they boarded the ship
* ALONE	Whether the person was alone or not.

LABEL:

• Given the above features we have to predict whether a person survived or not.
• survived(1) and didnt survive(0)

FEATURE SELECTION

• We have to remove the features that doesnt add value.
• We have to remove the features like EMBARK_TOWN as it is immaterial where a person boarded the ship
• We can add the features N_SIBLINGS_SPOUSES and PARCH since they are nearly the same.

MODELS USED:

• Logistic Regression
• Decision Tree
• Support Vector Machine
• Random Forest
• Boosting(using decision tree)
• MLP Classfier

MODEL SELECTION:

• Random Forest was selcted as the best model based on the F1_score.
• One drawback of Random Forest is that it is a bit slow
• If our aim is for faster training of the model, we can select the Decison Tree
• Decision Tree provides a good F1_score while the time taken to train is fastet than that of Random Forest.