Authors:

Parth Kapila

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

The Girl Scouts of Central Indiana relied on static, historical averages for forecasting, resulting in significant demand mismatches and revenue loss. This project replaces that outdated approach with a fully automated machine learning pipeline that transforms the forecasting process into a real-time, data-driven process. To address these challenges, the team implemented a two-phase approach: (1) Automation & Infrastructure Integration and (2) Optimization & Performance Enhancement. In the first phase, the team connected troop-uploaded sales data from Google Drive to a hybrid ML model built in Scikit-learn (Python), set up automated preprocessing, stored predictions in a PostgreSQL database, and deployed a web dashboard on Render to give troop leaders real-time access to forecast results. In the second phase, the focus will shift to improving model accuracy by tuning hyperparameters, experimenting with different algorithms, and refining the training process. The goal is to reach 90% accuracy, up from the current R² score of 0.877. Key outcomes demonstrated significant improvements in operational efficiency, model accuracy, and usability. The system is projected to save over $1.08 million in surplus reduction, equal to eliminating 1.35 excess cases per troop-cookie combination. This initiative offers a scalable blueprint for mission-driven organizations aiming to reduce waste, enhance forecasting accuracy, and enhance decision-making through responsible prediction and automation. Keywords: Automation; Machine Learning; SQL; Forecasting; Python

Source:

Purdue University / 2025

Topics:

No topics listed

Co-authors:

Parth Kapila