Font Size: a A A

Integration Model Construction And Land Use Scenario Simulation Based On Multi-layer Perceptron And Cellular Automata

Posted on:2020-07-11Degree:MasterType:Thesis
Country:ChinaCandidate:Y FengFull Text:PDF
GTID:2428330590971741Subject:Computer technology
Abstract/Summary:PDF Full Text Request
Land resources are not only the material basis of human production and life development,but also an important component of natural environment.While human being is developing continuously,it also influences and changes the changes of global ecological environment.In this thesis,the Three Gorges reservoir area was taken as the research area.The data set was composed of the land use status data of 2010 as the tag,13 factors and various land use neighbourhood factors together constitute 20 characteristics.Meanwhile,model was constructed to simulate the research area,and the model was used for scenario prediction.The main contents and conclusions are as follows:1.A model combining multi-layer neural network and cellular automata was constructed to simulate the land use situation in the Three Gorges reservoir area in 2010.By adjusting the network structure,the most suitable model is found.After many tests,the simulation accuracy and Kappa accuracy value were used as evaluation indexes.When the hidden layer was 3,the test accuracy of the model was the highest,reaching 0.959,and the overall Kappa accuracy value was 0.93.2.Optimize the model to solve the problem of unbalanced data samples.Data sampling method is used to process training data,and the raw data and processed data are respectively imported into the model combining neural network and cellular automata for learning.According to the test set,compared with the optimal network model before data processing,the total simulation accuracy is improved by 0.52%,and the best simulation accuracy was 0.964.The overall Kappa accuracy was 0.9372,which increased by 0.77%..The optimized model has significantly improved the simulation accuracy of the minority samples,among which the Kappa accuracy of shrubland type has been improved by 146.01%.The model was compared with convolutional neural networks,deep belief networks and CLUE-S.The results show that the model proposed in this thesis has some advantages over the other three models in precision.3.Different development scenarios are designed to predict the land use situation of the Three Gorges reservoir area in 2020,and the ecological service value index and economic benefit index are used to evaluate the quality of each development model.Finally,a multi-objective equation is established to simulate the land use situation with the optimal total benefit in 2020.4.Using Delphi and Python programming languages,a scenario simulation software combining deep neural network,cellular automata and oversampling algorithm is developed.
Keywords/Search Tags:Land use change, Neural network, Scenario simulation
PDF Full Text Request
Related items