🌠This work includes the program, Chinese comments, references, and a video on how to operate the program.🚀Software versions: MATLAB 2024b / MATLAB 2022a🌠Program acquisition method
Click the WeChat icon in the lower left corner to read the original text
💥Program test results displayThe simulation test results are as follows:
✨Algorithm Overview
The LSTM-based network traffic measurement process can be divided into core steps: model design, model training, model evaluation, and deployment application. Each step must be designed specifically considering the temporal characteristics of network traffic and the features of the LSTM model. The following details the objectives, operations, and mathematical modeling of each step. Based on the task objectives of network traffic measurement (such as traffic prediction and anomaly detection), the LSTM model’s network structure is designed, which includes “input layer → LSTM layer → fully connected layer → output layer”, while also selecting activation functions, dropout strategies, etc.
🪐PartProgram
...................................................................
figure;plot(IBN_testflow,'b');hold on
plot(pre,'r');grid on
xlabel('Time/s');ylabel('Traffic');hold on
plot(500*ones(1,300),10:10:3000,'k');text(100,2500,'Actual collected data and deep learning measurement results','fontsize',12);
text(600,2500,'Actual collected data and deep learning prediction results','fontsize',12);
hold on
plot(3000*ones(1,1000),'g--','LineWidth',2);
accacy = 100-100*mean(abs(IBN_testflow(1:end)-pre(1:end))./IBN_testflow(1:end));
legend('Actual IBN network traffic collected data','Deep learning IBN network traffic measurement data','','Threshold');
accacy title(['Deep learning traffic measurement accuracy:',num2str(accacy),'%']);
save LSTM.mat accacy IBN_testflow pre012_055m
