Getting results like this for all algorithms. Other than the MAPE, results are pretty consistent with other settings. Any ideas about why I’m getting such a large number for this? Implications?
You may check distribution of your target, possibly there are many zeros or close to zero values. If the raw returns (target) are skewed, the scaled data after min-max will be skewed as well. It is important to handle outliers properly.
If your target is between 0-1, and you use Rank then MAE can be nicely interpreted as absolute average deviation from percentile rank of target. There are many other reasons I like RANK the most.
1 Like