Groundwater is one of the primary sources of drinking water and is most vulnerable to contamination. This study applied quality assessment (health risk) indices, spatial dependence analysis, compositional data analysis (CoDa), and machine learning (ML) based clustering. The combination of spatial variability analysis, CoDa techniques, and Fuzzy c-means (FCM) clustering enhances the precision of identifying contamination zones, offering an effective alternative to traditional classification techniques that often fail to account for spatial and compositional relationships. Samples (n = 216) were acquired from a variety of sources and results findings revealed that groundwater contamination was under the respective limits set by WHO, except for TDS, Ca, Mg, and NO3 in 30 %, 46 %, 15 %, and 11 % of samples, while 88 % of samples were classified as fresh and very hard waters. The dominant hydro-chemical facies identified were Ca-Mg-HCO3 (78 %), followed by Ca-Mg-Cl-SO4 type (13 %). Ion exchange processes predominantly involved forward ion exchange across most of the area, while reverse ion exchange was more prevalent in the southeastern region of this study. The Van Wirdum diagram suggested that groundwater evolved from lithocline to munocline conditions. Notably, 7.5 % of samples pose health risks to children due to elevated NO3 levels in certain areas. CoDa identified anthropogenic activities as the primary contributors to increased K, NO3, and turbidity levels. Three distinct hydrogeochemical zones were delineated using FCM clustering, with notable NO3, pollution index of groundwater (PIG), and nitrate pollution index (NPI) hotspots in the southern part. ML-based clustering in the CoDa perspective is a useful tool for hydrogeochemical zonation.