Skip to main content Skip to secondary navigation

Predicting risk for ZUWA Solar Solutions

Main content start

Team members
Derek Betances
Brett Strickland
Jay Nagpal
Wakely Lush

Faculty mentor
Chuck Eesley

Sponsor organization
Zuwa Solar Solutions is a company dedicated to providing affordable and sustainable solar energy solutions to underserved communities. By leveraging innovative technology and financing models, Zuwa aims to bridge the gap in access to reliable energy sources, enhancing the quality of life and economic opportunities for its customers.

Project description

The primary objective of our project was to enhance Zuwa Solar Solutions' ability to predict customer default risk at the point of initial customer assessment. This predictive capability was crucial for optimizing the allocation of resources, tailoring financial solutions to customer needs, and ultimately expanding access to solar energy solutions in a sustainable manner.

Techniques and methods used

To achieve our goals, we embarked on a comprehensive data analysis and modeling project, employing various statistical and machine learning techniques. Initially, we explored multiple models, including linear regression, ridge and lasso regression, random forest, and gradient boosting, to identify the most effective approach for predicting customer default risk.

A critical phase of the project involved preprocessing the data, which included handling missing values, encoding categorical variables, and feature scaling. Despite the initial appeal of models like gradient boosting for their ability to handle multicollinearity and high variability, we discovered through experimentation that a random forest model provided the best balance between predictive accuracy and the practicality of data collection at the initial customer interaction.

Solutions and deliverables

The culmination of our project was the development of a refined random forest model tailored to Zuwa's operational context. This model was designed to utilize only data available at the initial customer assessment, ensuring its practical applicability without sacrificing significant predictive accuracy.

We delivered a comprehensive suite of tools and recommendations to Zuwa, including a prioritized list of variables that significantly influence default risk. This list not only informs Zuwa's data collection efforts but also serves as a basis for adjusting their customer risk assessment scorecard to reflect the predictive insights gleaned from the model.

Zuwa Solar Solutions plans to leverage our work to refine their risk assessment processes, enabling more accurate identification of default risks and thereby optimizing their resource allocation. 

By adopting a data-driven approach to customer assessment, Zuwa aims to expand its impact, providing sustainable solar energy solutions to even more households and communities. Our project's deliverables, including the predictive model and strategic recommendations for data collection and risk assessment, are set to play a pivotal role in enhancing Zuwa's mission of broadening access to renewable energy.

2024 senior projects