Adsorption prediction using an ANFIS may reduce the process costs in practical applications. MATLAB 2022 was used in this investigation to assess chromium (VI) adsorption data at different temperatures, doses of modified cellulose nanocrystals, and pH values. The best kind and quantity of membership function were chosen using the Box–Behnken experimental design approach to provide precise predictions with the least error. The RMSE was correlated with the number of MFs for each input by developing regression models based on analyzing every combination from the Box–Benkhen design. Five typical membership functions, Gaussian, triangle, Gaussian 2, trapezoid, and generalized bell-shaped, were used. ANOVA was used to demonstrate the significance of the regression models developed for the experimental data using trapezoid and trianglular MFs. The triangular MF produced the greatest accuracy in the Cr (VI) adsorption predictions (with a lower RMSE of 1.601 and R² of 0.998) when it was used in conjunction with the appropriate membership function numbers for each input (8, 8, and 4 for the trianglular membership function and 6, 6, and 3 for the trapezoid membership function) according to the ANFIS's predictions. Response surface plots were also used to evaluate the relationship between the triangular RMSE values and the membership function numbers. These findings demonstrate this material's potential to serve as a viable adsorption material for the focused elimination of contaminants, increasing the application of machine learning in sorption studies, and the remediation of novel pollutants.