摘要：On the basis of the leapfrog development oflcit- ies, the rise of industrial production, the advance- ment of scientific and technological means, and the increase in the total population have caused great damage to the ecological environment in the city, and the problems in the ecological environment have become increasingly obvious. The urban ecological environment is particularly fragile and sensitive andis greatly affected by human activities. Academic re-search on such issues is gradually increasing. A rea-sonable assessment of the sensitivity of the ecologi- cal environment in a city can help people better un- derstand it and grasp the main factors that are af- fected, to provide targeted solutions. Principal com-ponent analysis is used to analyze the driving forces that affect the fragility of the ecological environment in Daqing, including natural factors such as climate conditions and vegetation coverage, as well as social factors such as land development and population density. Then, based on the evaluation principle and the actual situation of Daqing, combined with the evaluation basis of principal component analysis, an index evaluation system is constructed. The BP artificial neural network method is used to evaluate the fragility of the ecological environment in Daqing. The research results show that according to the operating principles and rules of the BP neural network, the ecosystem vulnerability measurement framework of Daqing is constructed. After simula- tion training, the ecological environment vulnerabil- ity training results are 0.2045 for level I, 1.0045 for level II, and-0.1802 for level III. There is only anerror of 0.0051 between the output value and the ex- pected output value of level II. It is finally deter- mined that the ecological vulnerability of Daqing ismoderate. According to the error surface and contour line graph obtained from the research results, the to- tal variance data value of BP neural network training is relatively small, and the result is basically 0, indi- cating that the network training basically conforms to the expected goal.
