Most of the flooded layers in Chengdao oilfield are saline water flooded. Although the formation resistivity decreases monotonically with the increasing flooded degree，the relationship between the decline in formation resistivity and the flooding degree is extremely complex，and there is no effective method to identify the flooded layers and their flooding degree. Therefore，a prediction model of the flooded layer based on probabilistic neural network was proposed in this paper.Firstly，given the actual logging and test results of Chengdao Oilfield，the flooded degree was classified into five levels：unflooded，weak flooded，moderate flooded，strong flooded，and extra-strong flooded. The correlation analysis between logging characteristic parameters and flooding degree was carried out to select the characteristic parameters which could better reflect the flooding degree. Secondly，the extracted logging characteristic parameters and test results were employed to construct a learning sample library of the target probabilistic neural network model. Finally，the probabilistic neural network model was utilized to predict the flooded layers of the identified samples，and comparative analysis was conducted through the Adaboost algorithm that has good deep learning classification effects. The results show that the prediction accuracy of the water flooded layers is improved by10%，which improves the identification accuracy of the saline water flooded layers.