A comparison of artificial intelligence approaches in predicting discharge coefficient of streamlined weirs

Abstract In the present research, three different data-driven models (DDMs) are developed to predict discharge coefficient of streamlined weirs (Cdstw). Some machine-learning methods (MLMs) and intelligent optimization (IOMs) such as Random Forest (RF), Adaptive Neuro-Fuzzy Inference System (ANFIS), gene expression program (GEP) employed for prediction Cdstw. To identify input variables Cdstw by these DMMs, among potential parameters on Cdstw, most effective ones including geometric features weirs, relative eccentricity (λ), downstream slope angle (β), water head over crest weir (h1) determined applying Buckingham π-theorem cosine amplitude analyses. this modeling, changing architectures fundamental aforesaid approaches, many scenarios defined obtain ideal estimation results. According statistical metrics scatter plot, GEP model is a superior method estimate with high performance accuracy. It yields an R2 0.97, Total Grade (TG) 20, RMSE 0.032, MAE 0.024. Besides, generated mathematical equation in best scenario likened corresponding measured differences within 0–10%.