Skip to content

Built a Random Forest model and R Shiny dashboard to flag high-cost fleet assets and enable $800M+ savings for Reyes Holdings.

Notifications You must be signed in to change notification settings

sourabhgithubcode/predictive-fleet-maintenance

Repository files navigation

Predictive Fleet Maintenance – Reyes Holdings Capstone Project

This project was developed as part of a graduate-level capstone in Business Analytics at DePaul University. It focuses on using predictive analytics and clustering techniques to optimize fleet maintenance strategies for Reyes Holdings, one of the largest fleet operators in North America.

Objective

To identify high-cost maintenance assets and recommend data-driven strategies to reduce repair costs, improve asset utilization, and extend vehicle life using historical work order data.

Key Features

  • Built a Random Forest classification model to flag high-cost work orders (top 10% by total cost)
  • Achieved 98% prediction accuracy on testing data using engineered features like mileage ratio, asset age, and usage patterns
  • Developed clustering models to segment assets into strategy buckets: "Replace", "Monitor Closely", and "Maintain"
  • Created an interactive R Shiny dashboard for company-wide visibility and actionable insights (Link: https://6kk2xw-sourabh-rodagi.shinyapps.io/fleet_dashboard/)
  • Highlighted $800M+ in cost-saving opportunities based on asset-level segmentation

Tech Stack

  • Language: R (Tidyverse, randomForest, cluster)
  • Dashboard: R Shiny
  • Data Processing: dplyr, tidyr
  • Visualization: ggplot2
  • Data Size: 134,000+ fleet work orders with metadata, financials, and usage logs

Folder Structure

predictive-fleet-maintenance/ ├── data/ # Sample or cleaned datasets (anonymized) ├── scripts/ # R scripts for data cleaning, modeling, clustering ├── dashboard/ # R Shiny code and UI for final visualization ├── figures/ # Charts and exported graphs ├── README.md # Project summary └── report.pdf # Final business-facing capstone report

Impact

This project demonstrates the application of predictive analytics and operations strategy in a real-world enterprise fleet scenario. The results enable targeted interventions to lower maintenance spend and improve vehicle lifecycle management.

Author

Sourabh R Rodagi
M.S. Business Analytics, DePaul University
LinkedIn | Email

About

Built a Random Forest model and R Shiny dashboard to flag high-cost fleet assets and enable $800M+ savings for Reyes Holdings.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages