Ads final report - Team 8
- 1. Internet of Things Analytics ADVANCES IN DATA SCIENCE AND ARCHITECTURE TEAM 8 ANKITA SURESH SHANBHAG VIGNESH KARTHIKEYAN SRINIKETAN G.S
- 2. Problem Statement We have the weather station data from, for this analysis we have chosen Boston, MA and weather data from the year 2008 to present. We are going to forecast the weather for the next few days. Our Web application will display the forecast for the future period and will get updated every day. We are forecasting the future weather using the past data so we have used AR + I + MA modeling (ARIMA) for the forecast as the future data depends only on the past trend. And we have tried to observe the trend and seasonal components out of our data and the trend represents the gradual increase in temperature over the year, which might be due to global warming.
- 3. Column Type Timestamp Date EST character Max.TemperatureF numeric Mean.TemperatureF numeric Min.TemperatureF numeric Max.Dew.PointF numeric MeanDew.PointF numeric Min.DewpointF numeric Max.Humidity numeric Mean.Humidity numeric Min.Humidity numeric Max.Sea.Level.PressureIn numeric Mean.Sea.Level.PressureIn numeric Min.Sea.Level.PressureIn numeric Max.VisibilityMiles numeric Mean.VisibilityMiles numeric Min.VisibilityMiles numeric Max.Wind.SpeedMPH numeric Mean.Wind.SpeedMPH numeric Max.Gust.SpeedMPH numeric PrecipitationIn numeric CloudCover numeric Events character WindDirDegrees.br... character
- 5. It can be observed that there is a seasonality component to data but the trend is not clearly visible so on decomposing we get.
- 6. Detect the seasonality using Fourier transform
- 7. Code to get frequency The out put we get
- 8. Code to decompose The out put we get
- 9. Code to get (p,d,q) The out put we get Code to get forecast The out put we get
- 10. Import data
- 11. ARIMA Model
- 12. The forecast Data we get is:- (snap shot taken on 12 Dec 2016)
- 13. Add other weather components Weather component (p,d,q) Mean.TemperatureF (9,0,0) MeanDew.PointF (1,0,1) Mean.Humidity (2,0,2) Mean.VisibilityMiles (0,1,2) PrecipitationIn (0,0,1) CloudCover (0,0,1) Events Cannot be computed
- 14. forecast the weather for next 10 days
- 15. Events : compute Weather component (p,d,q) Variable Type Mean.TemperatureF (9,0,0) Independent MeanDew.PointF (1,0,1) Independent Mean.Humidity (2,0,2) Independent Mean.VisibilityMiles (0,1,2) Independent PrecipitationIn (0,0,1) Independent CloudCover (0,0,1) Independent Events Perform Regressio n Dependent
- 16. Filter out only the necessary variables
- 17. Score and Evaluate the model
- 18. Predict the ‘Events’ using dependent variable mentioned above
- 19. Forecast for 10 days into the future