AR model in MATLAB for short-term traffic flow prediction

1、 Content abstract

A little
438- Can communicate 、 consulting 、 Answering question

2、 Content description

The purpose of intelligent transportation system , Is to use rich traffic detection data , Predict the future traffic flow state , For the most efficient use of Expressway resources , Reduce traveler time , Reduce traffic congestion and traffic accidents on expressways . Short term traffic flow prediction is a road traffic control system 、 One of the most important problems to be solved in the field of traffic flow guidance system . Study the theory and method of short-term traffic flow prediction , So as to accurately predict the road traffic flow in minutes or even shorter time in the future , For alleviating urban traffic congestion 、 Avoiding the waste of social resources has important significance and application value . This paper takes the typical two adjacent intersections in the urban road network as the research object , Consider the traffic flow at the intersection in the past several times and the traffic flow related to the turning at the upstream intersection , A wavelet decomposition model is established to predict the first step , Combined with neural network, the hybrid prediction method for the second step prediction . By analyzing the characteristics of sample data of traffic flow at urban intersections , The short-term traffic flow shows strong nonlinearity 、 Time varying 、 uncertainty , And the correlation with the traffic flow at the upstream intersection . A single prediction method is difficult to meet the requirements of prediction accuracy , therefore , This paper proposes that the traffic flow data should be processed first “ frequency ” decompose , Firstly, wavelet transform is used to filter the traffic flow data , Filter the components of sudden changes in traffic flow caused by various uncertain factors into frequency bands , Get a basic data sequence and interference signal data sequence of different frequency bands , In this way, each group of data after decomposition becomes a relatively simple component 、 Data column with relatively stable data changes . Then use the model to predict the decomposed data in different frequency bands , Then add the prediction results of each sequence to get the first prediction . Finally, combined with the traffic flow of relevant turns at the upstream intersection, the neural network is used to predict the second step , Get the overall traffic flow prediction results with high accuracy .

key word ： short-term traffic flow forecasting ,AR Model , neural network , Mixed Forecasting

3、 Simulation analysis

clear,clc,close all
data = xlsread(' Four days 17 Point to the 20 Point traffic flow data .xlsx');
time = data(:,1);
flow = data(:,2);
num = length(time)/4;
figure
plot(time(1:num),flow(1:num))
title  The first day 17 Point to the 20 Point traffic flow data 
xlabel  Time /5min
ylabel  Traffic 
set(gcf,'color','w');
set(gca,'box','off');
figure
plot(time(num+1:num*2),flow(num+1:num*2))
title  the second day 17 Point to the 20 Point traffic flow data 
xlabel  Time /5min
ylabel  Traffic 
set(gcf,'color','w');
set(gca,'box','off');
figure
plot(time(num*2+1:num*3),flow(num*2+1:num*3))
title  On the third day 17 Point to the 20 Point traffic flow data 
xlabel  Time /5min
ylabel  Traffic 
set(gcf,'color','w');
set(gca,'box','off');
figure
plot(time(num*3+1:num*4),flow(num*3+1:num*4))
title  The fourth day 17 Point to the 20 Point traffic flow data 
xlabel  Time /5min
ylabel  Traffic 
set(gcf,'color','w');
set(gca,'box','off');

% y = iddata(flow(num*3+1:num*4),[],5*60);
% mb = ar(y,4,'burg')
% EstMdl = estimate(mb,time(num*2+1:num*3)); % y is your data
%%  The first day 17 Point to the 20 Point traffic flow data 
num = length(time)/4;
order = 6;
y1 = flow(1:num); y2 = flow(num+1:num*2); 
y3 = flow(num*2+1:num*3);  y4 = flow(num*3+1:num*4); 
model1 = ar(y1, order, 'burg', 'Ts', 5*60, 'IntegrateNoise', false)
p1_1=predict(model1,y2);
p1_2=predict(model1,y3);
p1_3=predict(model1,y4);
figure
plot(y2)
hold on
plot(p1_1)
hold off
figure
plot(y3)
hold on
plot(p1_2)
hold off
figure
plot(y4)
hold on
plot(p1_3)
hold off

 

4、 Reference paper

A little