Evaluate classifying with just year and max_depth

MetOffice / XBTs_classification

Project for the classification of eXpendable Bathy Thermographs

BSD 3-Clause "New" or "Revised" License

4 stars 2 forks source link

Evaluate classifying with just year and max_depth #54

Closed stevehadd closed 3 years ago

stevehadd commented 4 years ago

There is ome question about how much prdictability is coming from country and lat/lon vs max_depth and year. We should evaluate how much degradation in accuracy there is leaving out location info and just using year and depth as input features.

stevehadd commented 3 years ago

I have now created a basic notebook to show that results are reasonable and could use these features as a fallback option where country is not availabl. I have created a JSON experiment definition so this could be done. The results for this experiment should be generated as part of #46 to be included in the consolidated results notebook in #47 .