PoC for Air Quality models using machine learning for MWC 2018
- 1. 1 GSMA Internet of Things Seminar: Mobile World Congress 2018 Generating value through IoT and Big Data Dr. Hau Chen Mike Lee Far Eastone, Taiwan Case Study – Air Quality PoC and Transportation in Taiwan
- 2. 2 Air Pollution, an important living quality issue in Taiwan • 7 of top 10 causes of death in Taiwan, are closely related to air pollution. PM2.5 Taiwan Status 2
- 3. Universtiy of Texas. Environ. Sci. Technol. 2015, 49, 8057−8066 Harvard School of Public Health. Environ Health Perspect 2016 CVD & COPD Polluted Air Can Lead to Cardiovascular and Respiratory Diseases Increase 2.1ug/m3 of PM2.5, eGFR will decrease like the picture shown below accordingly.
- 4. 4 Air Quality IoT devices collected big data Governments/Companies install many connected devices to monitor Air Quality Values to Customers (Information) Qualities of sensors Quantities Qualities of data sets How often do we have data? How accurate? Can we use big data to forecast air quality? For how long? How good? DATA IoT Big Data AI/ Machine Learning 4 Forecast air quality
- 5. Objectives of “Air Quality Forecast PoC Project” 5 PM 2.5 NO2 Apply GSMA Greenwich Air Quality NO2 Forecast model to Taiwan. Add mobile network big data to improve the accuracy of NO2 forecast Expand to PM2.5 forecast
- 6. How we Built it Replicate GSMA IoT architecture & Machine Learning model in FarEasTone Acquire Taiwan air- quality & weather data into FET IoT harmonizer platform Verify GSMA London air-quality forecast model with Taiwan data Adding FET mobile network data & model tuning in Machine Learning model to improve forecast accuracy Architecture & ML Model Replication 1 TW AQ & Weather Data Acquisition 2 London ML Model Verification with Taiwan Data 3 4 Adding Mobile Network Data & Model Tuning in ML 6
- 7. 7 GSMA Greenwich NO2 Forecast Model NO2 Reading Temperature Pressure Tendency Wind Direction Weather Type Barometric Pressure Wind Speed Air Quality Data Weather Station Data Predict London N02 Value Gradient Boosting Regression Extra Trees Regression Random Forest Regression 1 2 3 4 NO2 _PPB Description NO2 _PPB Index Predicted NO2 ReadingNO2 _ Predicted PPB
- 8. Taiwan AQ & Weather Monitoring Stations in 3 Major Cities 6 Air Monitoring Stations 9 Air Monitoring Stations 1 2 3 4 5 Air Monitoring Stations Taipei Taichung Kaohsiung Select 5 weather stations nearby each air-quality station Weather StationAir-Quality Station 8
- 9. 9 Carrier Data & Model Tuning Helps NO2 Reading Air Quality Station Data Predict Taiwan N02 Value NO2 _PPB Description NO2 _PPB Index Predicted NO2 Reading Tempera- ture Wind Direction Barom- etric Pressure Wind Speed Weather Station Data Pollution Source Power Plant data FET mobile network data FET Population Aggregation 1 2 3 4 Gradient Boosting Extra Trees Random Forest Data Processing & Model Tuning Parameter Tuning Learning Rate, Max Features, Max_Depth Subsamples GSMA Model Result Grid Search Filter outlier Linear interpolation Data ProcessingOriginal NO2 _ Predicted PPB
- 10. 10 Summary of Model Verification Best Performance Model Base 9% 10% 13% 15% R2 Best Performance Increase Rate Predicted Taiwan NO 𝟐 GSMA Machine Learning Model Weather Data Interpolation FET Population Aggregation Power Plant Data Model Parameter Tuning (Grid Search) NO2 Reading = 0 Filtering NO2 FET Carrier Data ► Best Performance of Gradient Boost R2 : GSMA ML Model + Data Interpolation + FET Population + Power Plant Data + Model Tuning
- 11. 11 Mobile Network Data Improves AQ Forecast Top 10 Feature Importance Kaohsiung_Qiaotou Attribute Importa nce PopulationDensityDelta 2% PopulationDensity 2% SDamount3 2% NBamount2 2% HourOfDay 2% NO2Extrapolation 2% NO2D2 2% NO2D1 2% NO2Delta 2% AverageTemperature 2% Taichung_Dali Attribute Importa nce PopulationDensity 3% amount 2% PopulationDensityDelta 2% NO2D2 2% NO2Extrapolation 2% weatherC0F9A0Temperature D2 2% NO2Delta 2% NO2D1 2% HourOfDay 2% weatherC0F9A0Temperature D1 2% Taipei_Wanhua Attribute Import ance PopulationDensity 3% PopulationDensityDelta 2% NO2D2 2% HourOfDay 2% NO2Extrapolation 2% NO2D1 2% NO2Delta 2% AverageTemperature 2% weatherC0AH10Temperature 2% weatherC0AD30Temperature D2 2%
- 12. Forecast results for PM2.5 ► Besides NO2 forecasting, this model is also applicable to predict UK & Taiwan (Taichung& Kaohsiung) PM2.5 Britain_Loodo n GradientBoost Original ML R2 GR9 87% GR4 85% GN0 79% GN2 85% GN3 73% 12
- 13. 13 ► Asus Cloud deployed more than 250 airboxes in Taipei ► We can get more monitoring results than just government air quality stations AirBoxes We can provide reliable Air Quality Forecast on a per station level Taipei
- 14. What we achieved in PoC Air quality can be better monitored in Smart City, Health Care, and Industrial Anti- pollution by integrating : ► AI machine learning ► Carrier big data + IoT big data platform + AI model tuning ► Air Quality IoT Sensors GSMA's project has helped us developing a service that will be valuable to Taiwan AQ IoT sensor open data FET carrier data & IoT big data Health care Smart City Industrial Anti-pollution AQ forecast 14
- 15. We are in GSMA Innovation City 15 GSMA Innovation City The Air We Breathe
- 16. aisle GSMA Innovation City The Air we Breathe Please Come Visit and Share
- 17. 17 Thank you 17 FET is willing to share our development experiences with carriers