Micro-clustering and rank-learning profiling of a small water-quality multi-index dataset to improve a recycling process

Abstract

The efficiency improvement of wastewater recycling has been prioritized by ‘Goal 6’ of the United Nations Sustainable Development initiative. A methodology is developed to synchronously profile multiple water-quality indices of a wastewater electrodialysis (ED) process. The non-linear multifactorial screener is exclusively synthesized by assembling proper R-based statistical freeware routines. In sync with current trends, the new methodology promotes convenient, open and rapid implementation. The new proposal unites the ‘small-and-fast’ data-sampling features of the fractional multifactorial designs to the downsizing, by microclustering, of the multiple water quality indices—using optimized silhouette-based classification. The non-linear multifactorial profiling process is catalyzed by the ‘ordinalization’ of the regular nominal nature of the resulting optimum clusters. A bump chart screening virtually eliminates weak performances. A follow-up application of the ordinal regression succeeds in assigning statistical significance to the resultant factorial potency. The rank-learning aptitude of the new profiler is tested and confirmed on recently published wastewater ED-datasets. The small ED-datasets attest to the usefulness to convert limited data in real world applications, wherever there is a necessity to improve the quality status of water for agricultural irrigation in arid areas. The predictions have been compared with other techniques and found to be agreeable.