We have some trained models we want to use for anomaly detection. But training table had some columns that maybe arent present on prediction data. I mean, there are some parameters which appeared while training the models and now they arent in the data we need to use to predict anomalies. In this case, theorically we just need to manually create these parameters in the input (predict) table with zero values. But, is there any way to get the training table’s structure from these models ? Because we dont have these tables available right now, and we dont know how to get a list or similar of these missing parameters.
We also saved several models as pickled objects: python isolation forest, numeric outliers and pca models.