Titanic Classification

Overview

"This project analyzes the Titanic survival prediction dataset using machine learning techniques, focusing on data preprocessing, visualization, and model selection."

Dataset

• Source: Kaggle (Titanic-Dataset)
• Features: Passenger details (e.g., Sex, Age, Fare, Pclass)
• Target: Survived (0 = No, 1 = Yes)

Methodology

1. Load the Data: Imported Titanic dataset from Kaggle.
2. Explore & Visualize: Analyzed distributions and relationships (e.g., Sex vs. Survival).
3. Prepare Data: Handled missing values, encoded categorical variables, and scaled features.
4. Model Selection: Compared SGDClassifier, LogisticRegression, and RandomForestClassifier.
5. Conclusion: Selected tuned RandomForestClassifier (F1 = 0.7971).

Results

• Best Model: RandomForestClassifier with class_weight='balanced', max_depth=10, min_samples_split=5, n_estimators=300.
• Test F1-Score: 0.7971
• Precision: 0.8594
• Recall: 0.7432
• Key Features: Sex (0.361), log_fare (0.213), log_age (0.178).

Learnings

• Importance of handling imbalanced data with class_weight.
• Value of feature engineering (e.g., log transformations).
• Effectiveness of RandomForest for non-linear relationships.