Model Confidence Distribution

Accuracy76%

Recall (Critical Catch Rate)86%

F1 Score73%

Yielding massive friction recovery equating to ~$280K in modeled annual overhead savings.

IT Ticket Routing Analysis

Adidas IT • Gradient Boosting | Python • scikit-learn • Gradient Boosting • Classification Evaluation

The Problem

Helpdesk infrastructure costs were exponentially expanding because misrouted IT tickets suffered immense friction delays, driving massive opportunity and productivity loss.

The Methodology

Built an advanced Gradient Boosting classifier workflow around ticket urgency states, requested services, internal category tracking, and localized geography parameters.

The Impact & Outcome

De-risked the ticketing flow by constructing a functional AI classification filter to predict and flag high-risk routed tickets instantly at creation time.

Key Metric: 76% accuracy, 86% recall, 73% F1. Yielding massive friction recovery equaling roughly $280K in modeled annual overhead savings.

Classification Payload Distribution

Web browsers cannot securely render raw .pptx presentations globally without disrupting CSS page flows or requiring third-party PDF translation services. To view the Adidas Gradient Boosting presentation deck natively in Microsoft PowerPoint, please download the raw asset locally.

This artifact includes model evaluation charts, confusion matrices, and the complete ROC curves used in the final rollout evaluation.

Download Interactive .PPTX Deck