1 内容介绍
浮动车数据的挖掘是目前交通领域广泛使用的研究手段之一,基本的BP神经网络也多被用于交通流的预测研究.本文引入模糊算法,结合改进萤火虫算法寻优速度快,收敛率高的特点对基本的BP神经网络进行优化,构建了路网交通流速度的预测模型.利用城市路网浮动车真实数据对模型进行训练并通过测试数据对模型预测结果进行实证分析,证明模型对特定时刻路网交通流速度预测的准确性相比于基本BP神经网络算法有46.56%的提升,对路网24h内交通流速度预测的稳定性有39.08%的提升.
2 仿真代码
%% Well, Lets do it !!!
clc;
clear;
warning('off');
% Data Loading
data=JustLoad();
% Generate Fuzzy Model
ClusNum=4; % Number of Clusters in FCM
%
fis=GenerateFuzzy(data,ClusNum);
%
%% Tarining FireFly Algorithm
FireFlyFis=FireFlyRegression(fis,data);
%% Plot Fuzzy FireFly Results (Train - Test)
% Train Output Extraction
TrTar=data.TrainTargets;
TrainOutputs=evalfis(data.TrainInputs,FireFlyFis);
% Test Output Extraction
TsTar=data.TestTargets;
TestOutputs=evalfis(data.TestInputs,FireFlyFis);
% Train calc
Errors=data.TrainTargets-TrainOutputs;
MSE=mean(Errors.^2);RMSE=sqrt(MSE);
error_mean=mean(Errors);error_std=std(Errors);
% Test calc
Errors1=data.TestTargets-TestOutputs;
MSE1=mean(Errors1.^2);RMSE1=sqrt(MSE1);
error_mean1=mean(Errors1);error_std1=std(Errors1);
% Train
figure('units','normalized','outerposition',[0 0 1 1])
subplot(3,2,1);
plot(data.TrainTargets,'c');hold on;
plot(TrainOutputs,'k');legend('Target','Output');
title('FireFly Training Part');xlabel('Sample Index');grid on;
% Test
subplot(3,2,2);
plot(data.TestTargets,'c');hold on;
plot(TestOutputs,'k');legend('FireFly Target','FireFly Output');
title('FireFly Testing Part');xlabel('Sample Index');grid on;
% Train
subplot(3,2,3);
plot(Errors,'k');legend('FireFly Training Error');
title(['Train MSE = ' num2str(MSE) ' , Train RMSE = ' num2str(RMSE)]);grid on;
% Test
subplot(3,2,4);
plot(Errors1,'k');legend('FireFly Testing Error');
title(['Test MSE = ' num2str(MSE1) ' , Test RMSE = ' num2str(RMSE1)]);grid on;
% Train
subplot(3,2,5);
h=histfit(Errors, 50);h(1).FaceColor = [.8 .8 0.3];
title(['Train Error Mean = ' num2str(error_mean) ' , Train Error STD = ' num2str(error_std)]);
% Test
subplot(3,2,6);
h=histfit(Errors1, 50);h(1).FaceColor = [.8 .8 0.3];
title(['Test Error Mean = ' num2str(error_mean1) ' , Test Error STD = ' num2str(error_std1)]);
%% Plot Just Fuzzy Results (Train - Test)
% Train Output Extraction
fTrainOutputs=evalfis(data.TrainInputs,fis);
% Test Output Extraction
fTestOutputs=evalfis(data.TestInputs,fis);
% Train calc
fErrors=data.TrainTargets-fTrainOutputs;
fMSE=mean(fErrors.^2);fRMSE=sqrt(fMSE);
ferror_mean=mean(fErrors);ferror_std=std(fErrors);
% Test calc
fErrors1=data.TestTargets-fTestOutputs;
fMSE1=mean(fErrors1.^2);fRMSE1=sqrt(fMSE1);
ferror_mean1=mean(fErrors1);ferror_std1=std(fErrors1);
% Train
figure('units','normalized','outerposition',[0 0 1 1])
subplot(3,2,1);
plot(data.TrainTargets,'m');hold on;
plot(fTrainOutputs,'k');legend('Target','Output');
title('Fuzzy Training Part');xlabel('Sample Index');grid on;
% Test
subplot(3,2,2);
plot(data.TestTargets,'m');hold on;
plot(fTestOutputs,'k');legend('Target','Output');
title('Fuzzy Testing Part');xlabel('Sample Index');grid on;
% Train
subplot(3,2,3);
plot(fErrors,'g');legend('Fuzzy Training Error');
title(['Train MSE = ' num2str(fMSE) ' , Test RMSE = ' num2str(fRMSE)]);grid on;
% Test
subplot(3,2,4);
plot(fErrors1,'g');legend('Fuzzy Testing Error');
title(['Train MSE = ' num2str(fMSE1) ' , Test RMSE = ' num2str(fRMSE1)]);grid on;
% Train
subplot(3,2,5);
h=histfit(fErrors, 50);h(1).FaceColor = [.3 .8 0.3];
title(['Train Error Mean = ' num2str(ferror_mean) ' , Train Error STD = ' num2str(ferror_std)]);
% Test
subplot(3,2,6);
h=histfit(fErrors1, 50);h(1).FaceColor = [.3 .8 0.3];
title(['Test Error Mean = ' num2str(ferror_mean1) ' , Test Error STD = ' num2str(ferror_std1)]);
%% Regression Plots
figure('units','normalized','outerposition',[0 0 1 1])
subplot(2,2,1)
[population2,gof] = fit(TrTar,TrainOutputs,'poly4');
plot(TrTar,TrainOutputs,'o',...
'LineWidth',1,...
'MarkerSize',6,...
'MarkerEdgeColor','g',...
'MarkerFaceColor',[0.9,0.1,0.1]);
title(['FireFly Train - R = ' num2str(1-gof.rmse)]);
xlabel('Train Target');
ylabel('Train Output');
hold on
plot(population2,'b-','predobs');
xlabel('Train Target');
ylabel('Train Output');
hold off
subplot(2,2,2)
[population2,gof] = fit(TsTar, TestOutputs,'poly4');
plot(TsTar, TestOutputs,'o',...
'LineWidth',1,...
'MarkerSize',6,...
'MarkerEdgeColor','g',...
'MarkerFaceColor',[0.9,0.1,0.1]);
title(['FireFly Test - R = ' num2str(1-gof.rmse)]);
xlabel('Test Target');
ylabel('Test Output');
hold on
plot(population2,'b-','predobs');
xlabel('Test Target');
ylabel('Test Output');
hold off
subplot(2,2,3)
[population2,gof] = fit(TrTar,fTrainOutputs,'poly4');
plot(TrTar,fTrainOutputs,'o',...
'LineWidth',1,...
'MarkerSize',6,...
'MarkerEdgeColor','b',...
'MarkerFaceColor',[0.3,0.9,0.2]);
title(['Fuzzy Train - R = ' num2str(1-gof.rmse)]);
xlabel('Train Target');
ylabel('Train Output');
hold on
plot(population2,'r-','predobs');
xlabel('Train Target');
ylabel('Train Output');
hold off
subplot(2,2,4)
[population2,gof] = fit(TsTar, fTestOutputs,'poly4');
plot(TsTar, fTestOutputs,'o',...
'LineWidth',1,...
'MarkerSize',6,...
'MarkerEdgeColor','b',...
'MarkerFaceColor',[0.3,0.9,0.2]);
title(['Fuzzy Test - R = ' num2str(1-gof.rmse)]);
xlabel('Test Target');
ylabel('Test Output');
hold on
plot(population2,'r-','predobs');
xlabel('Test Target');
ylabel('Test Output');
hold off
%% Errors
% Fuzzy Regression Train and Test Errors]
% Train
fprintf('Fuzzy Regression Training "MSE" Is = %0.4f.\n',fMSE)
fprintf('Fuzzy Regression Training "RMSE" Is = %0.4f.\n',fRMSE)
fprintf('Fuzzy Regression Training "Mean Error" Is = %0.4f.\n',ferror_mean)
fprintf('Fuzzy Regression Training "STD Error" Is = %0.4f.\n',ferror_std)
fprintf('Fuzzy Regression Training "MAE" Is = %0.4f.\n',mae(data.TrainTargets,fTrainOutputs))
% Test
fprintf('Fuzzy Regression Testing "MSE" Is = %0.4f.\n',fMSE1)
fprintf('Fuzzy Regression Testing "RMSE" Is = %0.4f.\n',fRMSE1)
fprintf('Fuzzy Regression Testing "Mean Error" Is = %0.4f.\n',ferror_mean1)
fprintf('Fuzzy Regression Testing "STD Error" Is = %0.4f.\n',ferror_std1)
fprintf('Fuzzy Regression Testing "MAE" Is = %0.4f.\n',mae(data.TestTargets,fTestOutputs))
% FireFly Regression Algorithm Train and Test Errors
% Train
fprintf('FireFly Regression Training "MSE" Is = %0.4f.\n',MSE)
fprintf('FireFly Regression Training "RMSE" Is = %0.4f.\n',RMSE)
fprintf('FireFly Regression Training "Mean Error" Is = %0.4f.\n',error_mean)
fprintf('FireFly Regression Training "STD Error" Is = %0.4f.\n',error_std)
fprintf('FireFly Regression Training "MAE" Is = %0.4f.\n',mae(data.TrainTargets,TrainOutputs))
% Test
fprintf('FireFly Regression Testing "MSE" Is = %0.4f.\n',MSE1)
fprintf('FireFly Regression Testing "RMSE" Is = %0.4f.\n',RMSE1)
fprintf('FireFly Regression Testing "Mean Error" Is = %0.4f.\n',error_mean1)
fprintf('FireFly Regression Testing "STD Error" Is = %0.4f.\n',error_std1)
fprintf('FireFly Regression Testing "MAE" Is = %0.4f.\n',mae(data.TestTargets,TestOutputs))
3 运行结果
编辑
编辑
编辑
编辑
编辑
编辑
4 参考文献
[1]王昕, 黄柯, 郑益慧,等. 基于萤火虫算法-广义回归神经网络的光伏发电功率组合预测[J]. 电网技术, 2017, 41(2):7.
[2]黄新波, 宋桐, 王娅娜,等. 基于灰模糊萤火虫算法优化的变压器故障诊断方法:, CN103698627B[P]. 2016.
博主简介：擅长智能优化算法、神经网络预测、信号处理、元胞自动机、图像处理、路径规划、无人机等多种领域的Matlab仿真，相关matlab代码问题可私信交流。
部分理论引用网络文献，若有侵权联系博主删除。
