0

I've been thinking about how I can adapt the bike sharing regression example for ML.NET, see link at the bottom. If you look at the "DemandObservation" data structure used in this example you can see all the different fields used for the regression. My question is if it will still work even if I don't have all the data.

For example, maybe I have data like the day, month, year, etc, but I don't have weather or temperature. Will the regression algorithm still work if I just pass in blank values when i'm making a prediction, or will it cause predictions to become completely inaccurate? Would I have to pass in guess or previous values to get the algorithm to work? What i'm trying to do essentially is have the algorithm predict future bike sharing demand. Am I going down the wrong path with this approach? Thanks for your help in advance.

https://github.com/dotnet/machinelearning-samples/tree/master/samples/csharp/getting-started/Regression_BikeSharingDemand

Aubrey Russell
  • 11
  • 4
  • Usually you would handle missing data. Pandas is what's mainly used for Python. ML.NET has some transforms that can handle missing data. – Jon Feb 24 '19 at 00:00
  • Ok thanks for letting me know. I'm guessing its also not possible to predict multiple labels simultaneously with regression? – Aubrey Russell Feb 24 '19 at 00:33
  • Your features should be correlated somehow with your target. It could be direct or indirect. You don't have a temperature value but the temperature is correlated with the month. Maybe the year feature is correlated with economic inflation. And you can merge your data with other sources. Like a weather forecast web service. – Cihan Yakar Mar 08 '19 at 20:41

0 Answers0