Based on BP neural network model and system dynamics of the earth's ecological system network modeling

摘要

Nowadays, environmental problem of the whole earth attracts more and more attention from scientists, politicians and general people as well. What we need to know is not only the importance of a health environment, but also how the human activity influences our environment. In this paper, we apply network models to solve this problem. These models help us to know the exact influence of key factors to the earth's health through the data, and can provide reliable predicting of the future health. Also, by the help of other models, in this paper we explored the tipping point of the environment and designed the warning level. To predict the future health, the first model we used is BP neural network model, which is easy to operate, and as an advantage, it is easy to revise the key factors in this model, making it a flexible one when the relationships change among the key factors. However, the model can't show the relationships to us though it may take it into consideration. To improve the situation, we use model two, a more complex model based on system dynamics. This model requires a large amount of historical data. However, in this model, it is difficult to identify the variations which evaluate the health of the environment. To improve this, we combine the two models together, using the state variations in the system dynamic model as the input factors in neural network model, to obtain the output factors, which can be used to evaluate earth's health. Also, in order to find the tipping point, but not only use the coarse information obtained from the output factors to measure the earth's health, we use entropy as a standard instead. Entropy is influenced by radiation from the sun, photosynthesis of plants and human activity. It is a thorough measure of earth's health since it represents the order of whole system, but not only an aspect of it. What's more, the paper also considered the feedback loops in the environment in model 2. And we believe that our models can- provide extra assistance for decision makers to choose and use their policy.