Hi,
I love what you guys are doing. Keep on the good work.
Question:
Using Mindsdb + Mariadb I would like to get the probability for all items in a column with a limited set of values.
Lets take for example in the bank_churn example data. Suppose I train to predict the geography column.
Instead of getting the top country (e.g. Germany) I would like to get the probability for all countries
Germany 0.8
France 0.6
Italy 0.3
Etc…
Please advise.
Thanks!
Hi @nirbd, thanks for the kind words!
This is something that we have been working on, and it should be available on next week’s release. You can expect a follow up with instructions and a short example by then.
Cheers,
Patricio
Thanks Patricio. I’m waiting for it.
Hi @nirbd,
As of MindsDB 2.19.0, you can access this feature by specifying 'output_class_distribution': True in your predictor advanced_args dictionary.
Here’s a short example using the python interface:
...
predictor.learn(..., advanced_args={'output_class_distribution': True})
results = predictor.predict(...)
class_probs = results._data['{target}_class_distribution']
class_labels = results._transaction.lmd['lightwood_data']['{target}_class_map']
...
Let me know how it goes!
That’s amazing. Thanks!
Hi @nirbd, we are working on the AI tables integration. I’ll ping back with an example once it’s possible.
Hi,
Apologies for the delay. This information is now available in the {target}_explain field for the MySQL and Mongo APIs:
{"class_distribution": {
"5": 0.04799520883510369,
"6": 0.10778617414928329,
"7": 0.4064338499181312,
"4": 0.053223713685620955,
"3": 0.06277927869179267,
"8": 0.22103869370700427,
"9": 0.10074308101306396},
"predicted_value": "7",
"confidence": 0.69,
"prediction_quality": "confident",
"important_missing_information": [
"fixed_acidity",
"chlorides",
"free_sulfur_dioxide",
"total_sulfur_dioxide"]
}
thanks, looks great. Does it work also on MariaDB?
Indeed, it’s available on MariaDB
thanks
