FlintIO
A Good Place To Start: Titanic Survival Prediction Project
Author & Project Owner
Don C. SeverinAbout
This project is based on the dataset of Titanic passenger information, including demographic information such as age, gender, and fare class, as well as information on whether the passenger survived.
By working on this project, you will gain a deeper understanding of machine learning algorithms and their applications and develop your data analysis and model-building skills. Whether you are a beginner or an experienced data scientist, the Titanic prediction project is an exciting and challenging project that will help you enhance your data science skills.
Challenge
The main goal is to pre-process the data as well as you can, select the proper techniques, develop a model to predict whether someone survived the titanic disaster based on a the training set of people that survived or not and obtain a good score.
Solution
In a nutshell, our process was simple and contained three main steps to complete this project. Data Loading & Setup, Exploratory Data Analysis, Algorithm Selection this road map provided the sub-task necessary. We conducted data loading and setup by using Numpy and Pandas libraries to view and interact with the CSV files and VS code for environment creation to download required packages.
Exploratory data analysis (EDA) was the most time-consuming as there is always more to discover within the data set. However, we stuck to a simple approach, looked for negative and positive correlations, and made variated and bivariate visualizations that supported a passenger's reason for survival.
Finally, the algorithm selection came down to a few questions are we looking for a dimensionality reduction? If no, do we need a response from the system? If yes, are we predicting numeric values? Indeed we are. Will our algorithm prioritize speed over accuracy? Or visa versa. Then we could figure out the best types of classifiers we would introduce to the data.
Results
This project was fascinating and can have you sitting for hours just itching to add other methods like feature engineering, Model Selection, and Pre-processing algorithms to gain the best score. However, for the portfolio, we will periodically update the project and aim for better in other versions to come.
You can take a look at the source code with the link below.
7.9
Prediction Score on local server
7.0
Prediction Score on Kaggle
