<p>In western Mediterranean climates, meteorological factors serve as critical indicators of environmental conditions. For instance, elevated relative humidity reliably signals a coastal Mediterranean setting. In order to assess localized risks, such as those affecting human comfort and health, it is essential to prioritize the study of these key parameters or their derived metrics. Relative humidity is a significant factor in hygrothermal comfort modelling, alongside air temperature, wind speed, and others. However, it becomes the key indicator in coastal environments due to their consistently high moisture levels. By fixing relative humidity as the central input, we can more accurately quantify and anticipate days of thermal comfort or risk. This targeted approach ensures that the models remain both relevant to regional climates and effective for practical comfort assessments. This study aims to develops a multi-layer perceptron, based classifier (MLP-classifier) to predict hygrothermal comfort days in Tangier city, Morrocco. The model is optimized using genetic algorithms, Gaussian radial basis functions, different activation functions, number of neurons, and training algorithms. Inputs include relative humidity paired with temperature and vapor pressure deficit. The findings of this study demonstrate that a radial basis function network, when provided with relative humidity and temperature as inputs, attains over 98% accuracy and manifests exceptional stability. In comparison, our RBF classifier, using relative humidity paired with vapor pressure deficit, reaches 95% accuracy with similarly robust performance. The efficacy of the proposed methodology is substantiated by confusion matrices and visualization tools, which demonstrate a significant decrease in false negatives and an increase in correctly identified comfort days. Collectively, these findings substantiate the efficacy of each model input combination in reliably predicting hygrothermal comfort.</p>