Hi,

Extracting information from a mutli-category Bayesian model can become difficult when the model is built on a rather large data set and large variety of categories. Indeed, the read model component returns for each feature in the model a count and a normalized probability columns. If one wants to then filter out the most important features for each category, both scripting (looping through all properties) and unpivoting data are computive intensive.

Is there a different and more efficient to extract features and stats per category, for instance only those based on more than N count and with a normalized probaibility greater than p or lover than -p' ?

Extracting then filtering those from a 50k Mols and 1000 Assay IDs multi-category model takes... ages (>4 days) and returns over 300M records! Furthermore, an HTML table cannot display a single record prior unpivoting.

Cheers,

Pierre