Bengaluru Traffic Police
Clock It — before the city jams

Illegal parking
enforced where it
actually matters

Clock It turns 298,445 historical violation records into a bias-corrected enforcement schedule — telling each police station which junction to deploy to, at what time, and with what expected disruption reduction. No guesswork. No patrol habits. Just data.

Open dashboard View hotspot map
298k
Violations analysed
168
BTP-coded junctions
60.1%
Patrol-window bias corrected
0.98
Classifier AUC
Violations by hour of day — city-wide
Enforcement bias detected
60.1% of all violations fall between midnight and 6 AM. That is not when parking is worst. That is when patrol teams go out. Clock It corrects for this before scoring any junction.
Top junctions by Parking Disruption Index
PDI leaderboard — bias-corrected
The problem

Enforcement is reactive. Resources are finite. Congestion keeps growing.

Traffic police in Bengaluru operate on patrol habit and complaint-driven response. There is no system that tells them where illegal parking is causing the most disruption right now, and where a single team would have the highest impact.

Patrol-driven enforcement
Teams follow established routes. High-violation areas near patrol bases look critical on a raw heatmap. Low-patrol areas with worse parking go unnoticed.
Bias-corrected PDI scoring
Clock It divides violation density by patrol-window exposure for all 168 BTP-coded junctions, surfacing locations that are underenforced relative to actual demand.
Weekly enforcement schedule
A deployment calendar — which team goes where, at what time — that maximises city-wide disruption reduction with the available officer headcount.
What the data shows
Three numbers that explain why this system exists
60.1%
of all violations recorded at night
The raw data looks like parking is a 2 AM problem. It is not. Patrol teams go out at night. The data follows the patrols, not the violations. Every analysis that ignores this is analysing patrol habits, not parking behaviour.
13.4%
of records carry two or more simultaneous offences
A vehicle parked on a main road junction blocking both a carriageway and a footpath is not the same enforcement problem as a scooter on a side street. Clock It weights compound violations at 1.5x in the PDI formula.
96
violations at Builders Junction — PDI score of 64.6
A junction with 96 total violations outscores junctions with 15,000 because 47% of its records are multi-offence main-road blockages at peak hours. Raw count ranking would never surface this. PDI does.
How it works

Four layers. No black boxes.

Every formula is auditable. Every score is reproducible from the raw CSV in a single pipeline run.

01
Bias correction
Violation counts are divided by estimated patrol-window exposure per junction. Locations with high night-patrol intensity are down-weighted relative to sites sampled more evenly across the day.
Enforcement bias corrected
02
PDI scoring
Each of the 168 BTP-coded junctions receives a Parking Disruption Index — a weighted sum of bias-corrected frequency, main-road rate, repeat consistency, multi-violation rate, and peak-hour share.
5-component formula
03
Deployment optimisation
A greedy schedule maximises total PDI reduction across available teams per shift. Filter by police station to see exactly which junctions that station should prioritise, on which days, in which windows.
Greedy LP optimiser
04
ML classifier
A logistic regression trained on 22 raw observation features predicts which junctions are chronic structural hotspots vs. seasonal patterns. AUC 0.981, F1 0.839, 5-fold CV 0.963. Retrain from the dashboard in one click.
AUC 0.981
The insight others miss

The 3 AM spike is not the problem

Every other analysis of this dataset will show a spike in violations between midnight and 6 AM and conclude that is when enforcement should be heaviest. That spike is a recording artifact. It shows when patrol teams go out, not when parking is worst.

Clock It is the only system that corrects for this. The result is an opportunity map — junctions that are underenforced relative to their actual violation demand, not relative to where patrol teams already go.

See the full analysis View corrected hotspots
Raw violation count by hour — all 298,445 records
12 AM 6 AM 12 PM 6 PM 11 PM
Night patrol window (0–6h) — 60.1% of records
Evening peak (17–20h)

Ready to see where enforcement is most needed?

Open the dashboard to explore the full PDI leaderboard, deployment schedule, and ML model — all computed from the raw violation data.

Open dashboard View enforcement schedule ML model metrics