Help in insights for data

Hello, relatively new to KNIME and I need help/suggestions on any findings I can make with my data set
Dataset: My dataset is hourly from December 2010 to December 2011. My data is based on Beijing and measures its PM2.5 values, which is basically the air pollution level. Listed below are the variables in my dataset.

  1. Year: Year of data in the row
  2. Month: Month of data in the row
  3. Day: Day of data in the row
  4. Hour: Hour of data in this row
  5. Pm2.5: PM2.5 concentration (ug/mA3)
  6. DEWP: Dew Point
  7. TEMP: Temperature
  8. PRES: Pressure (hPa)
  9. Cbwd: Combined wind direction
  10. Iws: Cumulated wind speed (m/s)
  11. Is: Cumulated hours of snow
  12. Ir: Cumulated hours of rain
    And listed below are Hypothesis/Questions I have answered using Knime
  13. PM2.5 levels are higher on weekdays than on weekdays(Done)
  14. Which hour of the day are PM2.5 levels the highest?(Done
  15. Is PM2.5 levels directly/indirectly proportional to the other variables(Done)
  16. How frequently do PM2.5 levels go above 35(Unhealthy)(Done)
  17. How much does wind direction affect PM 2.5?(Done)
  18. What range of temperatures have the highest PM2.5 levels(Done)
  19. Which season in China is the best time to visit(Less polluted)(Done)
    8 Setting up linear regression model to predict PM2.5
    If anyone can give me suggestions on how I can further analyze this dataset and come up with more insights. Any help is greatly appreciated

If you had data from multiple PM monitors, the next thing would be to looks at how levels vary spatially, but it sounds like you probably don’t have that.

If you had data from PM sources, in theory you could use some sort of simple downstream Gaussian dispersion model to predict ground level impacts (although this would need a bit more sophisticated met data as well) - but I guess you don’t have that either.

In short, based on the data here, you’re probably going to be limited to the diagnostic evaluation of PM levels you’ve already done, rather than be able to predict them with any certainty.

EDIT: Is this for a course you’re taking? If so, I’d be interested to hear where, if you’re comfortable sharing that. :slight_smile:

1 Like