Abstract: Due to poor quality of seismic data,rare well log data and low degree of development in ML oilfield,the hydrocarbon prediction is not satisfactory by using conventional impedance inversion.The multi-attribute probabilistic neural network can take advantage of pre-stack and post-stack seismic data,and it is applied for seismic attribute technique to enhance the seismic data utilization ratio.Firstly,the multi-attribute analysis has been carried out to optimize the amplitude envelope,poisson's ratio and some other seven attributes.And then,probabilistic neural network algorithm is used to establish the nonlinear relationship between seismic attributes and hydrocarbon,and after that the distribution of hydrocarbon at the drilled sand has been predicted.It has been proved the prediction results are reliable for comparing the forecast results and the actual logging data.Finally,the distribution of hydrocarbon for the target sand body has been predicted and the probability and the thickness of the oil distribution can be achieved so as to provide guidance for the exploration and the development of lithologic reservoirs.The teleconology has achieved good effects in study area,and it is also worthy of learning for exploration and exploitation to the similar lithologic reservoirs.