DATA SCIENCE | Urban Delivery Demand Prediction | 40%

[liajohn30]

Reason:
Sprint 1 pull request – initial implementation of delivery demand prediction use case

Changes made:

• Created personal workspace in Playground
• Added delivery demand prediction notebook
• Performed exploratory data analysis (EDA)
• Applied data preprocessing and normalisation
• Implemented LSTM baseline model
• Generated initial visualisations (actual vs predicted demand)
• Added preliminary model evaluation

Notes:
The model currently captures general trends but does not fully capture demand spikes. Further improvements will include feature engineering and model tuning.

[Litxinh123]

Hi Lia,

Solid start on your submission so far

A couple of things you could improve in the next round:

• Include your personal details based on the required template (Authored by, Duration, Level, Pre-requisite Skills), please refer to the attached example
• Add more explanation after your code blocks to better justify your approach and highlight key takeaways

Happy to approve for now. Thanks!

[PriyanshuCauleechurn]

Overall:
Nice work on the notebook. The workflow is easy to follow, and I like that the preprocessing, baseline LSTM model, and T-GCN model are all organised in a logical sequence. It gives a clear picture of the full pipeline from raw delivery data to prediction.

What was done well:
The preprocessing steps are broken down clearly, which makes the notebook easier to read.
Good feature engineering choices such as time-based features and delivery duration
Nice progression from a baseline model to a more advanced graph-based model
The inclusion of evaluation, zone-wise analysis, and residual plots adds useful interpretation to the results.

Suggestions for improvement:
It may help to add a short markdown explanation before major sections, especially around why certain preprocessing choices were made, such as using KMeans for zones or choosing a sequence length of 6.

It may also help to include a short final comparison summary between LSTM and T-GCN, highlighting not just performance metrics but also what each model captures better.

Also, it would be great if in future merges you could include your personal details as mentioned earlier based on the provided template. (Author, duration, pre-requisite skills etc...)

Conclusion:
Overall, this is a strong notebook with a clear structure and good modelling direction. Happy to approve. Keep up the good work!