Explainable Machine Learning Prediction of PCB-138 Behavior Patterns in Edible Fish from Croatian Adriatic




Abstract:
Fish consumption, especially consumption of oily marine species, is globally increasing since it has been recommended by dieticians due to high content of polyunsaturated ω-3 and ω-6 (PUFAs) fatty acids in fish tissue. Health benefits of PUFA ingestion coincide with the risk of intake of hazardous lipophilic persistent pollutants including organochlorine pesticides (OCPs) and related polychlorinated biphenyls (PCBs). We examined the impacts of 18 fatty acids (FAs) and 36 toxic organic and inorganic contaminants on the behavior patterns of indicator congener PCB-138 in marine fish using eXtreme Gradient Boosting (XGBoost), SHapley Additive exPlanations (SHAP), and SHAP value fuzzy clustering. XGBoost indicated non-linear relationships between investigated variables that surpasses indications suggested by commonly applied correlation matrices. Ten extracted fuzzy clusters of SHAP values revealed that higher intake of saturated myristic-C14:0 and margaric- C17:0 acids followed by intake of nutritionally beneficial eicosadienoic acid (C20:2n-6) mostly contributed to the PCB-138 bioaccumulation. Important impacts on PCB-138 behavior patterns were also registered for chemically allied indicator congeners (-153 and -180) and organochlorines’ metabolite p,p’-DDE. Less prominent were the associations between target congener and the most toxic dioxin-like PCBs.

CITATION:

IEEE format

A. Stojić, B. Mustać, G. Jovanović, “Explainable Machine Learning Prediction of PCB-138 Behavior Patterns in Edible Fish from Croatian Adriatic,” in Sinteza 2020 - International Scientific Conference on Information Technology and Data Related Research, Belgrade, Singidunum University, Serbia, 2020, pp. 23-28. doi:10.15308/Sinteza-2020-23-28

APA format

Stojić, A., Mustać, B., Jovanović, G. (2020). Explainable Machine Learning Prediction of PCB-138 Behavior Patterns in Edible Fish from Croatian Adriatic. Paper presented at Sinteza 2020 - International Scientific Conference on Information Technology and Data Related Research. doi:10.15308/Sinteza-2020-23-28

BibTeX format
Download

RefWorks Tagged format
Download