Building an End-to-End Data Science Project in R

Preparing and Cleaning the Student Performance Dataset

This project uses the clean file from Project 2, so run 02_intermediate_project.R first. Save your code as 03_advanced_project.R.

The code loads the data, creates a Pass or Fail label from the average mark using a threshold of 60, and splits the students into training and test sets.

Keep the test set untouched until the end, so the final score is honest.

Performing Exploratory Data Analysis and Predictive Modeling

Now build the models:

• Linear regression to predict the average mark.
• Decision tree to predict Pass or Fail.
• Random forest to predict Pass or Fail.
• K-means clustering to find student groups.

We compare more than one model because one model alone can give a false sense of confidence.

Presenting Final Results Through Visualizations and a Project Report

The last part draws two charts and saves them. Then use report.Rmd, shown earlier, to share the results.

Share the scripts, data, and charts so anyone can repeat your work.

Main Results from the Project

• Model performance: On unseen students, the linear regression model achieved an R-squared of 0.806, explaining about 81% of the variation in marks in the test set. The RMSE was 3.99 marks, while the MAE was 3.08 marks.
• Study hours and variable importance: Each additional study hour was associated with about 5.3 additional marks in the fitted regression model, holding attendance constant. The random forest ranked study hours as the most important variable and section as the least important.
• Model comparison: The decision tree achieved 87.5% accuracy, while the random forest achieved 85% accuracy. This difference represents just one student out of the 40 in the test set, so cross-validation would help provide a more reliable comparison.
• Student clusters: Cluster 2 had the highest average study time (4.9 hours) and marks (68.5). Cluster 3 had the highest average attendance (87.2%) but the lowest average marks (53.9). The mean silhouette score of 0.338 suggests that the clusters are only moderately separated and overlap to some extent.

Note: These results come from synthetic practice data, in which study hours were designed to be a major driver of marks. They demonstrate how to interpret model outputs, but they should not be treated as conclusions about real students.

The complete project files for this tutorial are available on GitHub. The repository includes the R scripts, datasets, visualizations, and report files used throughout the Beginner, Intermediate, and Advanced levels.

GitHub Repository: https://github.com/hariniavanthikaaaaaa747/r-for-data-science

To run the projects, download the files, extract the project folder, and open it in RStudio. Follow the accompanying Screenshot and Run Guide for the complete instructions.

You can now transform data with advanced dplyr tools, build and test prediction models, and group data with clustering. You can write clean, safe functions and share results through R Markdown and Shiny.

You have also finished a full data science project from start to end. Keep practising on real public datasets to grow further.

Congratulations on completing the R for Data Science course!

You've come a long way, from learning R basics to building your own data science project. Keep practising with real datasets and exploring new techniques to sharpen your skills.

Ready to take the next step? Explore HCL GUVI's Data Science Course to learn more and gain hands-on project experience.

Keep practising with public datasets, experiment with different models, and build projects that showcase your skills. Every project you complete brings you one step closer to becoming a confident data science practitioner.

Keep learning, keep building, and keep growing!