Round 143

Round UUID: b46baf29-856f-4ce3-8043-01fe8ed3aa46

Prompt:

Classify coral reefs as 'Healthy' or 'Stressed' using these strict, explicit rules:

1. Classify as 'Stressed' ONLY if at least TWO of these major indicators are MODERATE or worse:
   - PollutionLevel: 'Moderate' or 'High'
   - ReefIntegrityScore: 'Moderate', 'Poor', or 'Very Poor'
   - AlgalCoverage: 'Moderate', 'Heavy', or 'Extensive'
   - DistantStressIndicators: 'Possible' or 'Confirmed'
   - BleachingEventsPerYear: 'Occasional', 'Frequent', or 'Constant'
   - PreviousStressIncidents: 'Minor', 'Moderate', 'Major' or worse
   - PredatorActivityLevel: 'Moderate High', 'High', or 'Severe'
   - CurrentFlowQuality: 'Fair' or 'Poor'

2. Classify as 'Stressed' if ANY combination of the following co-occur:
   a. BleachingEventsPerYear is 'Constant' AND at least one other major indicator (above) is MODERATE or worse.
   b. AlgalCoverage is 'Heavy' or worse AND at least one other major indicator is MODERATE or worse.
   c. DistantStressIndicators is 'Confirmed' AND at least one other major indicator is MODERATE or worse.

3. Classify as 'Healthy' ONLY if ALL these are TRUE:
   - ReefIntegrityScore: 'Good' or 'Excellent'
   - PollutionLevel: 'Low'
   - AlgalCoverage: 'Minimal' or 'Sparse'
   - BleachingEventsPerYear: 'Rare' or 'Occasional'
   - DistantStressIndicators: 'None' or 'Possible'
   - CurrentFlowQuality: at least 'Fair' (not 'Poor')
   - PredatorActivityLevel: not 'High' or higher
   - PreviousStressIncidents: not 'Moderate' or worse

4. For all other cases, if only ONE major indicator is Moderate or worse, classify as 'Healthy,' unless it is 'Confirmed' DistantStressIndicators or 'Constant' BleachingEvents AND all other major indicators are at their healthiest.

5. Do NOT classify as 'Stressed' based on ONE severe value alone unless described above in rule 2.

6. If criteria for neither are fully met, default to 'Stressed' if at least two are moderate/worse; otherwise, default to 'Healthy.'

Apply these rules strictly; no heuristics or subjective weighting. Use only the listed features and precise category definitions for each label.

• Inferences: 1000
• Created: 2025-04-17 02:05:13+00:00
• First inference: 2025-04-17 02:05:19+00:00
• Last inference: 2025-04-17 03:43:24+00:00
• Completed: yes
• Train accuracy: 0.31800766283524906 (KT 0.497)
• Validation accuracy: 0.3942307692307692 (KT 0.404)
• Test accuracy: 0.23893805309734514 (KT 0.621)