Instructions to use ayanjawed/transitpulse-planning-model with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Scikit-learn
How to use ayanjawed/transitpulse-planning-model with Scikit-learn:
from huggingface_hub import hf_hub_download import joblib model = joblib.load( hf_hub_download("ayanjawed/transitpulse-planning-model", "sklearn_model.joblib") ) # only load pickle files from sources you trust # read more about it here https://skops.readthedocs.io/en/stable/persistence.html - Notebooks
- Google Colab
- Kaggle
Add TransitPulse planning model and model card
Browse files- README.md +66 -0
- planning_model.joblib +3 -0
README.md
ADDED
|
@@ -0,0 +1,66 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
library_name: sklearn
|
| 3 |
+
tags:
|
| 4 |
+
- tabular-regression
|
| 5 |
+
- time-series
|
| 6 |
+
- transportation
|
| 7 |
+
license: other
|
| 8 |
+
---
|
| 9 |
+
|
| 10 |
+
# TransitPulse planning model
|
| 11 |
+
|
| 12 |
+
Predicts **hourly bus ridership for one route** (riders per route-hour) from information known in advance:
|
| 13 |
+
route, date, hour, schedule and weather. Built for transport planners as a CT-378 Machine Learning course project.
|
| 14 |
+
|
| 15 |
+
- Live demo (Hugging Face Space): https://huggingface.co/spaces/ayanjawed/TransitPulse
|
| 16 |
+
- Code, data builder and notebooks: https://github.com/ayanmirza99/ML-Project
|
| 17 |
+
|
| 18 |
+
## Intended use
|
| 19 |
+
- What-if planning for the 24 MTA bus routes it was trained on, for dates from 2025-01-06 to 2025-09-28.
|
| 20 |
+
- **Not** for other routes, other cities or dates outside that period (see Limitations).
|
| 21 |
+
|
| 22 |
+
## Model
|
| 23 |
+
`sklearn.pipeline.Pipeline`: one-hot encoding of `bus_route` and `borough`, then `HistGradientBoostingRegressor`
|
| 24 |
+
(learning_rate 0.05, max_iter 500, max_leaf_nodes 63). Selected by validation MAE among models whose saved file is under 50 MB;
|
| 25 |
+
a Random Forest was about as accurate but about 110 MB.
|
| 26 |
+
Input columns: bus_route, borough, gtfs_n_stops, gtfs_max_stops_per_trip, sched_trips_hr, sched_headway_min, sched_trips_day,
|
| 27 |
+
temperature_2m, precipitation, snowfall, wind_speed_10m, hour, day_of_week, is_weekend, hour_sin, hour_cos, is_am_peak, is_pm_peak,
|
| 28 |
+
is_holiday, is_school_day.
|
| 29 |
+
|
| 30 |
+
## Data (three sources plus calendar)
|
| 31 |
+
1. **MTA Bus Hourly Ridership: Beginning 2025**, NY Open Data (dataset id `gxb3-akrn`), summed to one row per route per hour.
|
| 32 |
+
Licence: the dataset page gives no licence information; it is covered by NY Open Data's terms of use.
|
| 33 |
+
2. **MTA GTFS static bus schedules** (mta.info/developers). A 2026 snapshot, not the 2025 service that produced the ridership.
|
| 34 |
+
3. **Open-Meteo historical weather** (New York City coordinates, one value per hour for the whole city).
|
| 35 |
+
4. US federal holidays (Python `holidays` package) and a hand-typed, approximate NYC school calendar.
|
| 36 |
+
|
| 37 |
+
Eight two-week windows between January and September 2025, 24 routes, 64,512 route-hour rows. Missing hours are filled with 0 riders.
|
| 38 |
+
Data are from New York because Karachi has no public passenger-count data.
|
| 39 |
+
|
| 40 |
+
## Evaluation
|
| 41 |
+
Time-ordered split by whole days: train 2025-01-13 to 07-14, validation 07-15 to 09-04, test 09-05 to 09-28 (17 days, scored once).
|
| 42 |
+
The final model is refit on train plus validation before testing.
|
| 43 |
+
|
| 44 |
+
| | MAE (riders) | RMSE | R2 |
|
| 45 |
+
|---|---|---|---|
|
| 46 |
+
| Planning model (this model) | 65.4 | 113.3 | 0.92 |
|
| 47 |
+
| Baseline: training average for the route, hour and weekday/weekend | 78.0 | 129.4 | 0.89 |
|
| 48 |
+
|
| 49 |
+
The average route-hour in the test period has 450 riders, so the error is about 14.5%.
|
| 50 |
+
A separate history model (uses recent ridership) reaches MAE 39.0 but needs recent data, so it is not deployed.
|
| 51 |
+
Other checks: leave-one-window-out MAE 53.0 vs baseline 68.9; unseen routes MAE 111 vs baseline 132.
|
| 52 |
+
|
| 53 |
+
## Limitations
|
| 54 |
+
- Only 24 routes and eight 2-week windows in one year; the test set is only 17 days, so small differences are not meaningful.
|
| 55 |
+
- Some routes changed level during the year (Q58, Q10, Q65 about 35% lower; Q25 about 32% higher from summer). The model does not know about service changes; its error on these routes is about three times larger.
|
| 56 |
+
- Weather helps a little, but partly because it hints at the season. It is not evidence that weather causes ridership changes.
|
| 57 |
+
- Weather was one city-wide value, and the test used recorded (not forecast) weather, so real use will have higher error.
|
| 58 |
+
- Schedule data is a 2026 snapshot; the school calendar is approximate; licence wording was not independently checked.
|
| 59 |
+
- New York data only; no claim is made about Karachi.
|
| 60 |
+
|
| 61 |
+
## How to load
|
| 62 |
+
```python
|
| 63 |
+
import joblib
|
| 64 |
+
model = joblib.load("planning_model.joblib") # needs scikit-learn 1.9.1
|
| 65 |
+
```
|
| 66 |
+
The input row must contain the columns listed above (see `features.py` in the project repository).
|
planning_model.joblib
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:829a2f16c03588bc1709920885e7a0d61bf1e159a166fe0b917350be109f1e0d
|
| 3 |
+
size 1533426
|