Summary
This paper describes development of a web-based expert system employing Naive Bayes classification to support objective grading of Arabica coffee at a processing facility in Bener Meriah Regency, Indonesia. Six attributes—screen size, bean colour, aroma profile, moisture content, defect count, and cupping score—were modelled to assign coffee to one of four quality categories, with posterior probabilities indicating classification confidence. Functional verification confirmed system logic integrity, though predictive accuracy against the held-out test set remains to be formally reported.
Regional applicability
This is a systems engineering application study from Indonesia with no direct relevance to United Kingdom farming or food systems. The methodology could be adapted for coffee quality assessment in any origin, but transferability depends on availability of labelled local grading data and alignment with UK import standards or specialty coffee sector needs.
Key measures
Posterior probability scores for quality classification; accuracy of Bayesian predictions against training dataset (80 records); functional test coverage (24 black-box scenarios, 12 white-box logic paths)
Outcomes reported
Development and functional verification of a web-based expert system using Bayes method to classify Arabica coffee into four quality categories based on physical and sensory attributes. System demonstrated 100% success rate on 24 black-box test scenarios and conformance on 12 white-box logic paths.
Supporting research
Bayesian decision-support system is explicitly applied to Arabica bean physical/sensory quality grading.
Limits: Functional code tests do not establish predictive accuracy; abstract explicitly states test-label accuracy remains to be established.
Topic tags
Dig deeper with Pulse AI.
Ask about this record, its theme or its relevance to UK farming and policy. Pulse AI uses selected catalogue evidence and cites the sources it draws on.