Final Project review - ML 12

[efantinatti]

Hi @hhsafa @Amankaushik2009 @clv-techlead @Tupay88

Project Feedback

Highlights

• Extensive narrative and business context. The README frames the Toronto Transit Commission’s streetcar delay problem and quantifies the potential business impact of better incident classification. Stakeholders, including operations teams, maintenance departments, planners, leadership and passengers are clearly enumerated.
• Comprehensive methodology. The team consolidates 12 years of raw data, performs EDA and feature engineering, then trains logistic regression, random forest and gradient boosting models. Detailed tables summarise per-class precision/recall and feature importance for the selected Gradient Boosting model.
• Actionable insights and recommendations. The results show the model’s strengths and weaknesses and highlight that “hour” and certain routes (501, 503) are the most important predictors. Business impact recommendations cover temporal resource allocation, route-specific investment and vehicle maintenance strategies.
• Reproducibility. The README includes a dedicated reproducibility section with two options: running notebooks or executing Python scripts. It outlines expected outputs and troubleshooting tips and references the requirements.txt file. The repository contains a requirements.txt file listing exact versions of required packages.
• Team organization. A detailed section describes team roles, workflow and collaboration process, and provides reflection video links for several members.

Opportunities for Improvement

• Repository hygiene. The project commits all 12 raw data files in data/raw, which makes the repository heavy and breaches DSI guidelines. Provide a script to download or fetch data instead of storing large files.
• Missing reflection videos. Reflection videos are provided for Hossein, Aman and Janice, but Antonio’s link appears to be missing and one video entry is empty.

The Checklist

• Narrative: OK – Thorough context and business motivation.
• Stakeholders: OK – Stakeholders and their motivations clearly identified.
• Methodology: OK – Detailed data consolidation, modelling and evaluation.
• Insights: OK – Provides actionable recommendations tied to model findings.
• Repo Hygiene: [CAN IMPROVE] – Raw data committed; heavy repository.
• Reproducibility: OK – Clear instructions with requirements.txt.
• Video: [PENDING] – Some reflection videos missing or placeholders.

Best,
E. Fantinatti

[efantinatti]

Hi @hhsafa @Amankaushik2009 @clv-techlead @Tupay88 just FYI there are still pending items.