Abstract: Adaptive pattern recognition has been applied to predict possible location of earthquakes and their magnitude in the Beijing-Tianjin area and its adjacent regions. It consists of discrimination between two pattern classes on the basis of binary attributes. A trial set of coefficients are progressively modified until maximum discrimination is obtained. The basic steps in the adaptive pattern recognition algorithm are as follows:1. Set power vector W to zero;2. Set a counter K to zero;3. In the epicentres of known earthquakes order first of all, compute the discriminant score for first epicentre area, using equation D₁ = WY₁;4. If this epicentre area belongs to group A go to 5, otherwise go to step 8;5. If D₁>0 go to step 11;6. Increase K by 1;7. Recompute the W from the formula W'=W+αY, and go to step 11;8. If D₁<0 go to step 11;9. Increase K by 1;10. Recompute the W from the formula W'=W -αY, and go to stcp 11;11. Repeat step 3-10 for second epicentre area and others;12. If K is zero then none of the epicentres areas are misclassified namely D_i = WY_i correctly discriminates between two groups;If K is positive, go to step 2.First, analysis is made of the geological features of historical and recent earthquake epicentres and are then grouped to 16 factors.Strong earthquake epicentres in Beijing-Tianjin area are classified into two pattern classes: epicentres of earthquakes with magnitude of 6.0-7.0 and those above 7.0.Then previously uncatagorized or unknown earthquake areas are classified into one of the two pattern classes, based on the discriminant function produced by the computer. The results then obtained from pattern recognition are noteworthy. The Tangshan earthquake occurred just between the two areas which after the calculation are shown to be the most dangerous zones.After the Tangshan earthquake another prediction was made of future earthquake location and their magnitude. No matter whether this prediction will be accurate or not, we still consider that adaptive pattern recognition may be a promising method for prediction of earthquake location and their time of occurrance and for the study of numerical prediction of earthquake or say predictive geology.

摘要： 本文应用自适应图象识别预报京津及邻区可能发生强震的地点和强度。自适应图象识别方法仅是根据二进制标志作两种图象等级的判别。此方法开始用一组试验系数逐步修改直到达到最大判别。在地震数值预报研究中,可能是一条有希望的途径。