This research employs ANN and RSM to create a system for removing lead and nickel ions from wastewater utilizing a chitosan derivative. Chitosan beads were formulated and subsequently grafted with 4-aminobenzoic acid to enhance binding sites. The adsorption investigations utilized pH, adsorbent dosage, contact duration, temperature, and concentration as input layer data. Simultaneously, two neurons functioned as output layers, signifying the adsorption of Pb2+ and Ni2+ ions. Statistical metrics, including average relative errors (ARE), coefficient of determination (R2), Marquart’s percentage standard deviation (MPSD), mean squared error (MSE), Pearson’s Chi-square (χ2), root mean square errors (RMSE), and sum of squared errors (SSE), were employed to assess the RSM and ANN models. The optimally trained neural network exhibits R2 values of 1.0, 0.968, and 0.961 for the training, validation, and testing phases. The findings indicated that the ANN strategy surpassed the RSM-CCD model approach. The optimization results of the RSM-CCD model for the process variables were achieved at pH 5, an initial concentration of 100 mg/L, an adsorbent mass of 6.0 g, a reaction duration of 55 min, and a temperature of 40°C, yielding maximum removal efficiencies of 99.43% for Pb2+ ions and 99.36% for Ni2+ ions. The data suggest that ANN can predict the elimination of contaminants from wastewater.