Groundwater aquifers are complex systems that require accurate lithological and hydrogeological characterization for effective development and management. Traditional methods, such as core analysis and pumping tests provide precise results but are expensive, time-consuming, and impractical for large-scale investigations. Geophysical well logging data offers an efficient and continuous alternative, though manual interpretation of well logs can be challenging and may result in ambiguous outcomes. This research introduces an automated approach using machine learning and signal processing techniques to enhance the aquifer characterization, focusing on the Quaternary system in the Debrecen area, Eastern Hungary. The proposed methodology is initiated with the imputation of missing deep resistivity logs from spontaneous potential, natural gamma ray, and medium resistivity logs utilizing a gated recurrent unit (GRU) neural network. This preprocessing step significantly improved the data quality for subsequent analyses. Self-organizing maps (SOMs) are then applied to the preprocessed well logs to map the distribution of the lithological units across the groundwater system. Considering the mathematical and geological aspects, the SOMs delineated three primary lithological units: shale, shaly sand, and sand and gravel which aligned closely with drilling data. Continuous wavelet transform analysis further refined the mapping of lithological and hydrostratigraphical boundaries. The integrated methods effectively mapped the subsurface aquifer generating a 3D lithological model that simplifies the aquifer into four major hydrostratigraphical zones. The delineated lithology aligned closely with the deterministically estimated shale volume and permeability, revealing higher permeability and lower shale volume in the sandy and gravelly layers. This model provides a robust foundation for groundwater flow and contaminant transport modeling and can be extended to other regions for improved aquifer management and development.