[BP prediction] BP neural network based on AdaBoost realizes data regression prediction with matlab code

1 brief introduction

BP  Network is a typical feedforward neural network , The method of error back propagation is used in the weight training , It has the approximation function of nonlinear continuous rational function . In the process of signal forward transmission , The input signal enters from the input layer , After hidden layer processing , Reach the output layer . The neuron state of each layer only affects the neuron state of the next layer . Judge whether the result of the output layer is the desired output , If not , Then turn to back propagation , Then adjust the network weight and threshold according to the prediction error , So that  BP  The predicted output of neural network keeps approaching the expected output . Because of its simple structure , There are many adjustable parameters , Many training algorithms , Good handling ,BP  Neural network has been widely used in practice .

Adaboost  The algorithm is  Boosting  One of the typical applications of the algorithm .AdaBoost The learning algorithm selects a small number of very important rectangular features to construct a series of weak classifiers , Then these weak classifiers are cascaded to form a strong classifier . The algorithm selects the rectangle feature that can best distinguish positive and negative samples . For every feature , The weak classifier gives the threshold of an optimal classification function , Make the least number of samples misclassified .Adaboost  The advantage of the algorithm is that it uses the weighted training data to replace the randomly selected training samples , Combine weak classifiers , Use weighted voting mechanism instead of average voting mechanism .​

2 Part of the code

%%  The code is based on BP_Adaboost The strong predictor of %%  Clear environment variables clcclear​%%  Download data load data1 input output​%%  Weight initialization k=rand(1,2000);[m,n]=sort(k);​% The training sample input_train=input(n(1:1900),:)';output_train=output(n(1:1900),:)';​% Test samples input_test=input(n(1901:2000),:)';output_test=output(n(1901:2000),:)';​% Sample weight [mm,nn]=size(input_train);D(1,:)=ones(1,nn)/nn;​% Training sample normalization [inputn,inputps]=mapminmax(input_train);[outputn,outputps]=mapminmax(output_train);​​%%  Results statistics % Strong separator effect output=at*test_simu;error=output_test-output;plot(abs(error),'-*')hold onfor i=1:8error1(i,:)=test_simu(i,:)-output;endplot(mean(abs(error1)),'-or')​title(' Absolute value of strong predictor prediction error ','fontsize',12)xlabel(' Prediction samples ','fontsize',12)ylabel(' The absolute value of the error ','fontsize',12)legend(' The strong predictor predicts ',' Weak predictor prediction ')​%%​

3 Simulation results

4 reference

[1] Li Xiang , Zhuquanyin . be based on Adaboost Algorithm and BP Tax forecasting based on neural network [J]. Computer application , 2012, 32(12):4.

About bloggers ： Good at intelligent optimization algorithms 、 Neural networks predict 、 signal processing 、 Cellular automata 、 The image processing 、 Path planning 、 UAV and other fields Matlab Simulation , relevant matlab Code problems can be exchanged by private letter .

Some theories cite network literature , If there is infringement, contact the blogger to delete .