Transparent open-box learning network provides auditablepredictions for coal gross calorific value

摘要

Auditing and forensic analysis of how each prediction is calculated are key attributes of transparent open-box learningnetworks (TOB). It provides the full calculation and input metric contributions for each of the predictions it derives. Thereare two stages in executing TOB predictions (stage 1 matches and ranks using squared-error analysis; stage 2 optimizes andconducts sensitivity analysis). Neither stage involves generating or extrapolating correlations between the input variables.Both stages of the calculation generate accurate predictions for datasets with multiple, highly-dispersed and non-linear influencinginputs. The transparent way in which generates predictions leads to better understanding of the interplays betweenthe input variables. Such attributes have direct relevance to the complex systems modelled in the coal industry [e.g., gascalorific value (GCV) prediction and coal petrology–grindability relationships]. The algorithm is applied here to predictGCA for a large published database (6339 records) of US coals including proximate and ultimate analysis metrics. TheTOB predicts GCV with accuracy (RMSE≤0.3 MJ/kg; R2>0.99). The transparency of the TOB method contrasts with thehidden relationships involved in many neural-network based prediction systems. Worked examples are provided to show thedetailed prediction calculations associated with individual data points. The TOB approach applied to predicting coal GCVcan help to verify the source of specific samples (e.g. specific mines or coal basins) using readily understandable underlyingcalculations available for audit and display. The TOB is therefore also suitable for identifying the provenance of specificcoal samples based on proximate and/or ultimate analysis.