基于Kaggle训练集预测的多层人工神经网络的能源消耗的时间序列预测（Matlab代码实现）

news2024/9/24 17:13:56

目录

💥1 概述

📚2 运行结果

🌈3 Matlab代码实现

🎉4 参考文献

💥1 概述

本文为能源消耗的时间序列预测，在Matlab中实现。该预测采用多层人工神经网络，基于Kaggle训练集预测未来能源消耗。

📚2 运行结果

部分代码：

%% Exercise 2 - Recurrent Neural Networks

%% Initial stuff
clc
clearvars
close all

%% 1 - Hopfield Network

% Note: for this exercise, it will be interesting to check what happens if
% you add more neurons, i.e. modify T such as N is bigger.

T = [1 1; -1 -1; 1 -1]'; % N x Q matrix containing Q vectors with components equal to +- 1.
% 2-neuron network with 3 atactors
net = newhop(T); % Create a recurrent HOpfield network with stable points being the vectors from T

A1 = [0.3 0.6; -0.1 0.8; -1 0.5]'; % Example inputs
A2 = [-1 0.5 ; -0.5 0.1 ; -1 -1 ]';
A3 = [1 0.5 ; -0.3 -0.4 ; 0.8 -0.6]';
A4 = [0 -0.1 ; 0.1 0 ; -0.5 0.1]';
A5 = [0 0 ; 0 0.1 ; -0.1 0 ]';
A0 = [1 1; -1 -1; 1 -1]';
% Simulate a Hopfield network
% Y_1 = net([],[],Ai); % Single step iteration

num_step = 20; % Number of steps
Y_1 = net({num_step},{},A1); % Multiple step iteration
Y_2 = net({num_step},{},A2);
Y_3 = net({num_step},{},A3);
Y_4 = net({num_step},{},A4);
Y_5 = net({num_step},{},A5);

%% Now we try with 4 Neurons

T_ = [1 1 1 1; -1 -1 1 -1; 1 -1 1 1]'; % N x Q matrix containing Q vectors with components equal to +- 1.
% 4-neuron network with 3 atactors
net_ = newhop(T_); % Create a recurrent HOpfield network with stable points being the vectors from T

A1_ = [0.3 0.6 0.3 0.6; -0.1 0.8 -0.1 0.8; -1 0.5 -1 0.5]'; % Example inputs
A2_ = [-1 0.2 -1 0.2 ; -0.5 0.1 -0.5 0.1 ; -1 -1 -1 -1 ]';
A3_ = [1 0.5 1 0.5 ; -0.3 -0.4 -0.3 -0.4 ; 0.8 -0.6 0.8 -0.6]';
A4_ = [-0.5 -0.3 -0.5 -0.3 ; 0.1 0.8 0.1 0.8 ; -0.7 0.6 -0.7 0.6]';
% Simulate a Hopfield network
% Y_1 = net([],[],Ai); % Single step iteration

num_step_ = 40; % Number of steps
Y_1_ = net_({num_step_},{},A1_); % Multiple step iteration
Y_2_ = net_({num_step_},{},A2_);
Y_3_ = net_({num_step_},{},A3_);
Y_4_ = net_({num_step_},{},A4_);

%% Execute rep2.m || A script which generates n random initial points and visualises results of simulation of a 2d Hopfield network 'net'

% Note: The idea is to understand what is the relation between symmetry and
% attractors. It does not make sense that appears a fourth attractor when
% only 3 are in the Target T
close all