fix(api): use zero confidence fallback on prediction error

[devprashant19]

Description

This PR fixes a bug in the /predict endpoint (backend/api.py) where the confidence score calculation silently swallowed exceptions and returned a random, hardcoded fallback percentage between 65% and 99%. This misled users by presenting a fake confidence level whenever the underlying model's predict_proba or decision_function threw an error.
Fixes #224

Changes

• Removed the random.uniform(65, 99) fallback in the except Exception: block around line 205.
• Set the fallback confidence to a deterministic and safe 0.0 (or None) when an exception occurs, accurately reflecting that the model's confidence could not be computed.

Related Issues

• Resolves the fake/random confidence score bug.

Type of Change

• Security fix
• Bug fix
• New feature
• Refactoring

Pre-submission Checklist

• Verified that exceptions during confidence calculation no longer generate misleading random values.
• Checked that normal prediction confidence scores are still accurately extracted from the model.

[vercel]

@devprashant19 is attempting to deploy a commit to the Aditya Sharma's projects Team on Vercel.

A member of the Team first needs to authorize it.