Abstract: A new method is described for predicting reservoir properties using seismic data based on sedimentary micro-facies and sand thickness features of ES2x in Yong 3 block of Yong'an area.Conventional reservoir predicting method is to cross-plot the target data and seismic attribute for deriving the desired relationship between the two,and has a low predictive precision.Probabilistic neural network (PNN) method uses the convolutional operator to resolve the frequency difference between seismic attribute and the log data.The reliability of reservoir predicting results can be checked by cross-validation.Cross-validation divides the entire training data set into two subsets:the training data set and the validation data set.The training data set is used to derive the transform,while the validation data set is used to measure its final prediction error.Validation error can be used to check the validity of the attributes transform.On the basis of above research,6 optimum attributes susceptive to porosity are selected.Predicting precision is appraised by calculating prediction error and validation error and by correlating the prediction porosity and actual porosity.The reservoir porosity of sand layer of ES2x in Yong 3 block of Yong'an area is predicted successively by using PNN at last.