Transformer-LSTM預(yù)測(cè) | Matlab實(shí)現(xiàn)Transformer-LSTM多變量時(shí)間序列預(yù)測(cè)

效果一覽

基本介紹

1.Matlab實(shí)現(xiàn)Transformer-LSTM多變量時(shí)間序列預(yù)測(cè)，Transformer結(jié)合LSTM長(zhǎng)短期記憶神經(jīng)網(wǎng)絡(luò)多變量時(shí)間序列預(yù)測(cè)；

2.運(yùn)行環(huán)境為Matlab2023b及以上；

3.data為數(shù)據(jù)集，輸入多個(gè)特征，輸出單個(gè)變量，考慮歷史特征的影響，多變量時(shí)間序列預(yù)測(cè)，main.m為主程序，運(yùn)行即可,所有文件放在一個(gè)文件夾；

4.命令窗口輸出R2、MSE、RMSE、MAE、MAPE、MBE等多指標(biāo)評(píng)價(jià)；

程序設(shè)計(jì)

• 完整程序和數(shù)據(jù)下載私信博主回復(fù)Matlab實(shí)現(xiàn)Transformer-LSTM多變量時(shí)間序列預(yù)測(cè)。

%%  清空環(huán)境變量
warning off             % 關(guān)閉報(bào)警信息
close all               % 關(guān)閉開(kāi)啟的圖窗
clear                   % 清空變量
clc                     % 清空命令行%%  導(dǎo)入數(shù)據(jù)
result = xlsread('data.xlsx');%%  數(shù)據(jù)分析
num_samples = length(result);  % 樣本個(gè)數(shù)
or_dim = size(result, 2);      % 原始特征+輸出數(shù)目
kim =  2;                      % 延時(shí)步長(zhǎng)（kim個(gè)歷史數(shù)據(jù)作為自變量）
zim =  1;                      % 跨zim個(gè)時(shí)間點(diǎn)進(jìn)行預(yù)測(cè)%%  數(shù)據(jù)集分析
outdim = 1;                                  % 最后一列為輸出
num_size = 0.7;                              % 訓(xùn)練集占數(shù)據(jù)集比例
num_train_s = round(num_size * num_samples); % 訓(xùn)練集樣本個(gè)數(shù)
f_ = size(res, 2) - outdim;                  % 輸入特征維度%%  劃分訓(xùn)練集和測(cè)試集
P_train = res(1: num_train_s, 1: f_)';
T_train = res(1: num_train_s, f_ + 1: end)';
M = size(P_train, 2);P_test = res(num_train_s + 1: end, 1: f_)';
T_test = res(num_train_s + 1: end, f_ + 1: end)';
N = size(P_test, 2);%%  數(shù)據(jù)歸一化
[P_train, ps_input] = mapminmax(P_train, 0, 1);
P_test = mapminmax('apply', P_test, ps_input);[t_train, ps_output] = mapminmax(T_train, 0, 1);
t_test = mapminmax('apply', T_test, ps_output);%%  數(shù)據(jù)平鋪
P_train =  double(reshape(P_train, f_, 1, 1, M));
P_test  =  double(reshape(P_test , f_, 1, 1, N));t_train = t_train';
t_test  = t_test' ;%%  數(shù)據(jù)格式轉(zhuǎn)換
for i = 1 : Mp_train{i, 1} = P_train(:, :, 1, i);
endfor i = 1 : Np_test{i, 1}  = P_test( :, :, 1, i);
end

參考資料

[1] https://blog.csdn.net/kjm13182345320/article/details/128163536?spm=1001.2014.3001.5502
[2] https://blog.csdn.net/kjm13182345320/article/details/128151206?spm=1001.2014.3001.5502