In this study， the best solid adsorbent for decolorization was first determined by single-factor test， and then the effects of decolorization temperature， solid adsorbent addition and decolorization time on the decolorization rate of fish oil were investigated. On this basis， the Box-Behnken(BB) experimental design was used to optimize the decolorization conditions of fish oil， and the results of the BB test were analyzed by response surface methodology (RSM) and artificial neural network (ANN). The results showed that the activated white clay had the best decolorization effect， and the decolorization rate tended to increase and then decrease with the increase of decolorization temperature， solid adsorbent addition and decolorization time； the correlation coefficient r， determination coefficient R2， root mean square error RMSE and mean square error MSE values of RSM and ANN models were 0.9647， 0.9307， 1.1000， 1.2100 and compared with the RSM model， the ANN model fits better and has less error between the measured and predicted values， so it is more suitable as a prediction model for the decolorization rate of fish oil. Therefore， the RSM and ANN models were selected to jointly optimize the fish oil decolorization process. Firstly， the optimal decolorization conditions were selected by the RSM model， with decolorization temperature of 93.79 °C， solid adsorbent addition of 4.80%， and decolorization time of 9.69 min. Subsequently， the above conditions were brought into the ANN model， and the maximum decolorization rate of fish oil was finally obtained as 99.53%. The results of this study showed that the RSM-ANN model has strong accuracy and applicability， which provides a new idea for process optimization in food research.