Hyderabad AQI Prediction System
Real-time AQI monitoring and forecasting for 15 Hyderabad stations, with five ML models, health advice and PDF reports; the system behind the AQI paper.
- Monitoring stations
- 15
- Across Hyderabad
- Prediction models
- 5
- Regression, LSTM, Random Forest, Transformer, deep net
- RMSE
- 3.7
- Multi-modal framework, from the paper
01
Problem
- Who
- Residents, city planners and policymakers in Hyderabad.
- Why it mattered
- Industry, traffic and weather create non-linear, time-dependent pollution patterns that traditional AQI forecasting misses.
- Before
- Single statistical models that miss non-linear and seasonal effects.
02
Architecture
- Component
- Key decision (click)
Ingest
- WAQI APILive readings from 15 stations
- Weather dataTemperature, humidity, wind, rainfall
Predict
- Linear regression
- Random ForestFeature importance + confidence
- TransformerHugging Face, WebGPU
- Deep network9 → 64 → 32 → 16 → 1
Explain
- Interactive chartsRecharts, comparative analysis
- Health recommendationsBy AQI level
Report
- PDF reportsjsPDF + html2canvas
Pick a station, a future date and time, trend and seasonal factors; the five models forecast AQI from historical and live data, and the results are charted, explained with health advice and exported as a PDF.
03
Key decisions
D1Several model families side by side
- Alternatives considered
- A single forecasting model
- Why
- Linear, sequence, tree and deep models capture different parts of the pollution signal, and comparing them shows where each is reliable.
04
Build highlights
- 01
15 stations, one map
New Malakpet, Nacharam, ECIL, US Consulate, Bahadurpura Zoo Park, Somajiguda, Sanathnagar, Kokapet, Central University, Patancheruvu, IDA Pashamylaram, IITH Kandi, Kompally, Ramachandrapuram and Bollaram.
- 02
Full pollutant picture
Tracks PM2.5, PM10, NO₂, O₃ and more, with trend indicators and weather conditions.
- 03
Backed by peer research
The multi-modal framework (LSTM, Transformer, Random Forest) is published in the International Journal of Research Publication and Reviews.
05
Results
3.7
RMSE
Multi-modal deep learning framework
94.65%
Accuracy
vs conventional statistical methods
+45%
Winter PM2.5
Above the yearly average
How this was measured: From the published paper: historical data from 15 Hyderabad monitoring stations, with real-time WAQI data.
06
Links & media
- Live producthyderabadaqi.netlify.app (opens in a new tab)
- Paper / researchRead the research write-up
- Not publicThe source code is private.
