MFS metadata usage as feature data

Pushback to the Future Challenge
1

Hello,

I read the instruction about which cases we are allowed/not allowed to use MFS metadata, but I’m unclear yet. I’d be very appreciative if you could elaborate more on this.

The problem description stated that we could use the metadata for the current flight for which predictions are generated. Can we also use the metadata as a feature for the training model?

2

Yes, you may use the metadata to construct training features, subject to the same constraints: it should only be used to lookup information from flights you already know about from other sources (this can be the flight you are predicting or a flight you know about from another feature). Please let me know if you have any other questions or need additional clarification!

1 Like
3

Can we use some heuristics rules/learning models to infer/predict the aircraft type/ aircraft engine class for the GUFIs that do not exist in the metadata?

4

Apologies for the delay; we wanted to make sure we had thought carefully about your question. I’d first refer to the section regarding MFS data in the Problem Description:

You may:

  • Look up (or “join”) metadata for GUFIs that you already know exist within the valid time window from other timestamped features.
  • Use MFS data for the current flight that predictions are being generated for.

You may not:

  • Use the MFS metadata to look for information about flights for which you do not already have a GUFI from timestamped features.
  • Analyze the entire MFS metadata file to, for example, directly incorporate the distribution of aircraft type, carriers, etc. into your solution.

If what you are thinking satisfies those constraints, then what you propose, creating heuristics to estimate MFS features (e.g., aircraft type, engine class) for GUFIs that are not in the metadata is allowed. Let me know if you have any questions about the specifics!