MACHINE LEARNING PROJECTS

OVERVIEW

This project is a famous competition project in Kaggle. It is the Titanic project. In this project, I analyzed data from survivors of the sinking of the Titanic ship. This project is quite simple, just create a model that can predict whether a passenger has a chance of surviving or not based on certain variables.

From the resulting plot, we can see it turns out that:

• More passengers departing from Cherbourg survived, while passengers from Southampton, only half survived.
• For people who only bring 1 Parent/Child, more survivors.
• For people with Passenger Class 1, the chances of survival were much higher and for passengers with Passenger Class 3, only a few survived.
• People who bring 1 Sibling / Spouse are much more likely to survive. People who bring 2 Siblings / Spouse' chances of survival are quite small.
  

‍

It can be seen on the graph that the more expensive the ticket, the more likely the person is to survive, starting at 56 and up. But the number of survivors starting at 10.5 has increased considerably, except for the incident that happened to the group of ticket holders costing 13–15,742.

Based on the graph, it can be seen that passengers in the younger age group tend to be less safe than children and the elderly. The passenger age group is dominated by the young age group, which is between the ages of 18–24, 24–30, and 30–37 years.

The plot explains that people with Family_Size 2,3,4 have a greater chance of surviving, the rest probability decreases, and people who go with Family_Size_Grouped Small is more likely to have a greater level of safety.

‍

In the plot results, the explanation is the Family_Size_Grouped feature, at numbers 2, 3, and 4 the probability of survival increases.

‍

The total sample on y_train and y_train_pred is 891 passengers. The results of the confusion matrix can be explained as follows:

• There were 492 passengers or 55.2% who on the label were actually un-survived passengers and were accurately predicted as un-survived passengers.
• There were 57 passengers or 6.39% who on the label were actually passengers who did not survive but were predicted to be survivors.
• There were 89 passengers or 9.98% who on the label was actually survived passengers but were predicted to be un-survived passengers.
• There were 253 passengers or 28.39% who on the label are actually survived passengers and are accurately predicted as survivors.

‍

Based on the pie chart above, there are 61.48% of passengers who are predicted not to survive and there are 38.52% of passengers who are predicted to survive.

‍