Machine Learning Approaches to Passenger Survival Prediction: A Titanic Dataset Analysis

Abstract: The Spaceship Titanic competition on Kaggle offers a unique machine learning challenge inspired by the classic Titanic disaster, reimagined in a futuristic space setting. The goal of this research is to create a predictive model that decides if passengers on the Spaceship Titanic were transported to another dimension after the ship collided with a spacetime anomaly. This study examines different data preprocessing, feature engineering, and classification techniques to improve predictive performance on the dataset.

The dataset provides details on passenger demographics, travel itineraries, and onboard service usage, offering rich information for building classification models. Our approach involves meticulous data cleaning, handling missing values through imputation, normalizing numeric features, and encoding categorical variables. We evaluated several supervised learning algorithms—Logistic Regression, Random Forest, Gradient Boosting, and XGBoost—using metrics like accuracy, precision, recall, and F1-score. The top-performing model demonstrated strong generalization on the test set, highlighting the value of analyzing feature interactions and fine-tuning hyperparameters. This research underscores the power of data-driven techniques in predictive analytics and illustrates how machine learning can effectively tackle complex, hypothetical problems with incomplete and noisy data. Our findings contribute to the broader understanding of structured data prediction and highlight the effectiveness of ensemble methods for classification in both practical and simulated scenarios.