Extraction of tracer elements of particulate matter emission source using association rule mining

Abstract

The source marker species represent the emission source in the ambient air. It aids in identifying specific emission sources, but it can deliver ambiguous results when similar species exhibit different emission sources. Therefore, robust marker/tracer species are always needed to clearly identify emission source. This study collected tracer elemental species for possible emission sources from Indian published literature and processed them by using machine learning-based apriori algorithm to obtain robust elemental marker of the emission source. Initially, significant rules were chosen with support >10% and lift >1.0. Subsequently, elemental markers have obtained by applying constraint i.e. conviction ≥1.1, lift ≥1.4 and confidence ≥20%. As an outcome, it reveals elemental markers of crustal emission (CE) (Al, Si, Fe, Ca, Mg, Ti), sea salt (SS) (Na, K), biomass burning (BB) (K), solid waste burning (SWB) (Ba, Cd, Cr, Sr), coal combustion (CC) (As, Se, Cr, Cd), oil combustion (OC) (V, Ni, S, As), and traffic emission (TE) (Cu, Pb, Zn, Mn, Cd, Ni). Finally, robust elemental markers corresponding to their respective emission source were derived by applying constraint of conviction ≥1.1, lift ≥1.5 and confidence ≥50% on the previously extracted association rule. Consequently, it defined CE by Mg, Al, Ca, Si, Fe, SS by Na, K, TE by Cu, Pb, Mn, Zn, SWB by Ba, Cr, Cd, Sr, OC by V, Ni, S, and CC by Se, As. Additionally, this study also demonstrate the successful implementation of the apriori algorithm for the aforementioned task.