This repository contains a complete workflow for predicting passenger survival on the Titanic dataset using machine learning.
The project is structured into three main parts: Exploratory Data Analysis (EDA), Data Preprocessing, and Model Building.

• titanicEda.ipynb – Exploratory Data Analysis (EDA) of the Titanic dataset, including:

  • Data overview and summary statistics
  • Missing value analysis
  • Feature distributions and correlations
  • Insights for feature engineering

• DataPreprocessing.ipynb – Data cleaning and preprocessing steps:

  • Handling missing values
  • Encoding categorical variables
  • Feature scaling/normalization
  • Splitting into train/test sets

• model.ipynb – Machine Learning model training and evaluation:

  • Multiple algorithms (e.g., Logistic Regression, Random Forest, XGBoost)
  • Model comparison
  • Hyperparameter tuning
  • Performance metrics (accuracy, precision, recall, F1-score, ROC-AUC)

• Python 3.x
• Jupyter Notebook
• Pandas, NumPy – data manipulation
• Matplotlib, Seaborn – data visualization
• Scikit-learn – preprocessing & machine learning models
• RandomForestClassifier – Random Forest model
• Logistic Regression – Logistic regression model
• XGBoost – Ensemble Method

Logistic regression Cross Validation performance

• Accuracy: 0.807 (±0.020)
• Precision: 0.732 (±0.031)
• Recall: 0.789 (±0.071)
• F1: 0.757 (±0.034)
• Roc_auc: 0.863 (±0.024)

Random Forest Classifier Cross Validation performance

• Accuracy: 0.802 (±0.034)
• Precision: 0.754 (±0.046)
• Recall: 0.722 (±0.068)
• F1: 0.736 (±0.051)
• Roc_auc: 0.856 (±0.041)

XGBoost Cross Validation performance

• Accuracy: 0.810 (±0.026)
• Precision: 0.753 (±0.047)
• Recall: 0.760 (±0.036)
• F1: 0.755 (±0.028)
• Roc_auc: 0.857 (±0.028)

Logistic Regression seems to slightly outperform Random Forest Classifie However, Random Forest might be the better choice for precision (minimizing false positives)

Best Parameters: {'max_depth': 10, 'min_samples_leaf': 2, 'min_samples_split': 2, 'n_estimators': 100}

Tuned Random Forest Cross Validation performance

• Accuracy: 0.829 (±0.027)
• Precision: 0.786 (±0.036)
• Recall: 0.766 (±0.075)
• F1: 0.773 (±0.044)
• Roc_auc: 0.869 (±0.037)

Key insights:

📌 Future Improvements

Deploy as a Flask/Streamlit web app for interactive predictions

Add automated hyperparameter optimization (GridSearchCV/Optuna)

Experiment with deep learning models (TensorFlow/Keras)