Predicting Titanic Passenger Survival with RStudio: Random Forest & Logistic Regression using Tidymodels

Project Summary

This project explores machine learning techniques for predicting the survival of passengers aboard the Titanic. The primary focus is on applying Random Forest and Logistic Regression models to the Titanic dataset using the tidymodels package in R. The Titanic dataset is a well-known dataset that contains information about passengers, including whether they survived or not, and various attributes such as age, class, sex, and the number of siblings/spouses aboard.

The project aims to compare the performance of both models and evaluate their effectiveness in predicting passenger survival.

Key R Markdown File

This project contains an important .Rmd file that demonstrates moeddling using tidymodel. You can view the R Markdown file here: ML Model Markdown file.

Objective

The objective of this project is to:

• Build two predictive models (Random Forest and Logistic Regression) to predict passenger survival on the Titanic.
• Use the tidymodels package in R to create, train, and evaluate the models.
• Compare the performance of both models using appropriate evaluation metrics.
• Analyze the results and draw conclusions about the models' effectiveness in predicting Titanic survival.

Study Plan & Steps

Step 1: Data Exploration and Preprocessing

• Load the Titanic dataset from the titanic package or directly from a CSV file.
• Examine the dataset to understand its structure, missing values, and outliers.
• Perform data cleaning: Handle missing values, outliers, and any necessary transformations on categorical variables (e.g., encoding 'Sex', 'Embarked').
• Feature Engineering: Create new features, such as family size age, and possibly remove redundant or irrelevant features.

Step 2: Data Splitting

• Split the dataset into training and testing sets using a 80/20 split.
• Create resampling strategies for model validation, such as cross-validation using the vfold_cv function.

Step 3: Model Building & Training

• Random Forest Model:

  • Use rand_forest() from tidymodels to define and train the Random Forest model.
  • Tune the hyperparameters using grid search or random search.

• Logistic Regression Model:

  • Use logistic_reg() from tidymodels to define and train the Logistic Regression model.

Step 4: Model Evaluation

• Evaluate the models using appropriate metrics such as accuracy, and AUC.
• Compare the performance of the two models using the testing set and resampling methods.
• Visualize the confusion matrix and ROC curves for both models.

Step 5: Interpretation and Conclusion

• Analyze the importance of the variables in both models, especially for the Random Forest model.
• Discuss which features contribute most to the survival prediction.
• Evaluate the results in terms of model performance and interpretability.

Conclusion & Next Steps

Conclusion:

• Both models demonstrated the ability to predict Titanic survival, with Random Forest potentially providing higher accuracy due to its flexibility with non-linear relationships between features.
• Logistic Regression is simpler and more interpretable, which is useful for understanding how individual variables contribute to predictions.

Next Steps:

• Further hyperparameter tuning could be explored for Logistic Regression models to optimize their performance.
• Investigate other machine learning algorithms (e.g., Support Vector Machines, K-Nearest Neighbors) to compare with the current models.
• Apply advanced feature engineering techniques to enhance model performance, such as interactions between features.
• Implement model performance monitoring to track real-time performance if the model is deployed in production.

Technologies Used

• R programming language
• tidymodels package for machine learning
• randomForest for Random Forest modeling
• ggplot2 for data visualization

Dependencies

• tidymodels
• titanic
• ggplot2
• randomForest
• dplyr