Abstract：In horizontal reservoir prediction using seismic attributes，engineers face two problems:how to choose suitable seismic attributes，and how to analyse the reliability of prediction result. A simple method for solving above problems is developed by analysing some factors causing false correlation probability. This method achieves seismic attribute optimization and sample learning in three aspects:choosing some independent seismic attributes by correlation analysis，seismic attribute optimization and detailed output category which are based on samples，and typical sample optimization based on statistical analysis. The method is characterized by simple principle and easy realization. The application examples in TZl and TZ6 borehole areas show that its prediction results are more reasonable than those of other ordinary methods.
