Abstract: n PL oilfield of Bohai Sea Area,the density of the crude oil in Neogene Minghuazhen Formation is high,and the conventional poststack hydrocarbon detection methods are not effective in these oilfields.On the basis of the principle that frequency energy will be absorbed and attenuated if there is hydrocarbon in the reservoirs,the energy coefficient method and the attribute of the slope attenuation of high frequency have been used for predicting hydrocarbon with high precision spectral decomposition technique,and the forecast results are not good enough.Combined with the geological characteristic of the study area,we select optimal seismic attributes as training parameters by using a variety of neural network algorithm based algorithm,which have different responses for oil and water.Using oil saturation curve as a target curve,we have conducted seismic inversion for the whole study area.We can know the coincidence rate of oil sands in main reservoir reached above 75%between the verification wells and the saturation volume.By using these methods,combined with study results on the geologic condition for hydrocarbon accumulation,several sets of sand bodies with good hydrocarbon shows have been selected as potential further exploration targets.