In this study, an Adaptive Neuro-Fuzzy Inference System (ANFIS) model is proposed for the determination of alpha radioactivity of Hazar Lake waters and for the prediction of its unknown values. The model parameters of the lake water are pH, total hardness (TH), depth, electrical conductivity (EC), and alpha radioactivity. ANFIS model is performed using the back-propagation algorithm, which has the five layers. Average relative error between measurements predicted by theoretical (ANFIS) and experimental data is approximately 0.7043%. The relative error between the test data and the radioactivity data change between 0.06% and 14%. Additionally, validity of the model is also tested with a regression model. The predicted results with the ANFIS model is better as statistically than the regression model.