Predictive Capability of Phosphate Recovery from Wastewater Using a Rough Set Machine Learning Model
P. Balasubramanian, Muhil Raj Prabhakar, Chong Liu, Fayong Li, Xuan Cuong Nguyen, Nageshwari Krishnamoorthy, Zipeng Zhang, Pengyan Zhang
Abstract
Phosphate recovery from wastewater treatment plants has recently become the focus of the waste management sector. Indeed, a real-time recovery facility on a commercial scale necessitates using advanced prediction techniques to improve planning and execution procedures. Although many machine learning studies have been reported to predict phosphate recovery, there are no general rules for recovering phosphate from different wastewater sources. This study aims to apply rough set machine learning (RSML) to predict phosphate recovery based on decision attributes. This rule-based classifier model generates IF–THEN rules to classify the conditional attributes to achieve a proper decision. The model has identified the solution pH, initial phosphate (P) concentration, and Mg/P and Mg/N ratios along with operational parameters like reaction time and retention time as core attributes; oversight of these attributes may lead to misprediction of P recovery. The model generated 104 rules, including 16 approximate rules. The accuracy of trained RSML was 91.42%, performing better than that of the existing classifier models. The results of this RSML-based predictive model will have significant implications on scientific rules for future work in the fields of wastewater management and phosphate recovery.