Results change depending on the water quality evaluation methods used, and within good-quality water, many results still have parameters with concentrations exceeding the World Health Organization (WHO) desirable limits or national threshold values (TVs). Furthermore, there are few methods to classify the severity degree of contaminated water; most methods have problems in the parameter threshold boundary and in assigning weights. Aiming to solve the above problems, a water quality evaluation framework based on the single-indicator evaluation method (SIE), Weber–Fechner (W-F) law and Probabilistic Neural Network (PNN) is presented, named SIE&W-F&PNN. Forty-three confined water samples were collected for this research in Xi'an in September 2015. The SIE, water quality index (WQI) with three different weights (method weight, entropy weight and equal weight), comprehensive evaluation method (CEM) and SIE&W-F&PNN method were used, and the evaluation criteria for contaminated water were proposed based on the W-F law. The results of these methods were compared. The reasons for confined water pollution in Xi'an were analyzed. The results show that TC, NH4-N, NO2-N, β, As, Mn, F−, TH, Fe2+ and Turb were the contaminating parameters of the 43 confined water samples. In order, the results for the number or ratio of 'Poor' and even worse water samples by method are as follows: SIE-WHO (30, 69.77%) > SIE-GB = CEM (24, 55.81%) > WQI (entropy weight) (12, 27.91%) > WQI (method weight) (10, 23.26%) > WQI (equal weight) (9, 20.93%). These discrepancies highlight the influence of evaluation methods on the results. For this study, a water sample was classified as 'contaminated (bad) water' if any parameter exceeded either the national TV or the WHO's desirable limit, prioritizing drinking water safety. The SIE&W-F&PNN results show that there were 10 excellent water samples and 33 bad water samples (among which 4 water samples were rated as VL (very lightly polluted), 14 as L (lightly polluted), 14 as M (moderately polluted) and 1 as H (heavily polluted)). The SIE&W-F&PNN method ensures that no parameters in 'excellent' or 'good' water samples exceed the WHO's desirable limits or national TVs; can be used to classify the severity of contamination of contaminated water without assigning weights, avoiding the rate mutation near the threshold boundary; and can include any number of parameters and be applied to lakes, rivers, air, soil, etc. (i.e., it is not unique to groundwater). The primary causes of confined water pollution in Xi'an include historical pollution, contemporary anthropogenic activities, geological factors, excessive groundwater extraction, and the infiltration of contaminated surface and phreatic water. [ABSTRACT FROM AUTHOR] more...