Titanic: Machine Learning from Disaster

This repository contains my prediction solution for the Kaggle Titanic competition. I used the Snowflake ML Classification model to tackle this challenge, both to learn and to evaluate its efficiency.

Table of Contents

• Project Overview
• Getting Started
  • Prerequisites
  • Download the Data
• Running the Prediction
• Submit Prediction
• Score
• Acknowledgements

Project Overview

The Titanic competition on Kaggle is a classic machine learning challenge. The goal is to build a predictive model that answers the question: "what sorts of people were more likely to survive?" using passenger data (i.e., name, age, gender, socio-economic class, etc.).

I described the entire solution step-by-step in this article.

Getting Started

Prerequisites

To run the code in this repository, you need to ensure the following prerequisites:

• Create a Kaggle account.
• Install the Kaggle API and configure Kaggle API credentials. Instructions are available here.
• Ensure you have an active account on the Snowflake platform. If you don't have one yet, sign up for a Snowflake account on their website.

Download the Data

Once you have the Kaggle API and credentials configured, you can download the competition data by running:

kaggle competitions download -c titanic

Unzip the data into the data folder, remove the original zip archive, and you are ready to go:

unzip titanic.zip -d data/
rm titanic.zip

Running the Prediction

To run the prediction, follow the steps:

1. Create a new Snowflake Database (in my case it's called KAGGLE).
2. Load the train and test data sets into the tables (in my case they are named titanic_train and titanic_test).
   • You can do this through the Snowflake Web UI by uploading files through the web browser or by loading them from an S3 bucket. You can use the these SQL queries for loading the data that were generated when I was loading the data. Make sure to replace the filename.
   • The data schemes of the tables with train and test data should look like this.
3. Run one of the sequences of the queries:
   • Basic Classification Prediction
   • Classification Prediction based on some custom features and cleaned data
4. Save results to a .csv file that only contains two cloumns:
   • PassengerId
   • Survived

Submit Prediction

You can submit your score through the Kaggle Web UI or with the Kaggle CLI tool. The following command can be used to do it:

kaggle competitions submit -c titanic -f {file} -m "{message}"

Score

The advanced prediction with no some features engineering scored 0.78229, meaning it successfully predicted for ~78.23% of people whether survived or not.

I finished in 2810th place out of 17,290 at the time of submitting the results, which is in the 84th percentile.

Maybe I'll keep working on it and see if I can improve the score with some feature engineering, or alternative approach.

Acknowledgements

Thanks ❤️ to Kaggle for providing the Titanic dataset and competition. It is a super fun and educational challenge.