forecasting method in matlab using VAR model

Good evening sir

I request you please resolve the error from my code. I attached DATA also sir .

CODE:

% Step 1: Load data

data = xlsread('DATA.xlsx');

time = data(1:400, 1); % Time column (1st column)

inf_data = data(1:400, 2); % Infected data column (2nd column)

% Step 2: Prepare data for VAR

% Since we are using univariate data, we treat it as a single-variable time series

% Create lagged data (you can modify the lag length)

lags = 2; % Number of lags

dataForVAR = [inf_data(1:end-lags), inf_data(lags+1:end)];

% Step 3: Fit the VAR model

model = varm(2, 1); % 2 lags, 1 variable

fitModel = estimate(model, dataForVAR); % Fit the model

% Step 4: Forecasting using the VAR model

numForecasts = 10; % Number of steps to forecast

% Ensure 'Y0' is a numeric matrix, not a vector

presampleData = dataForVAR(end-lags+1:end, :); % The last lags data as presample

YForecast = forecast(fitModel, numForecasts, 'Y0', presampleData);

% Rescale the forecasted data (if needed)

forecastData = YForecast(:, 1); % First column is the forecast for the 'inf_data' series

% Step 5: Plot the results

figure;

hold on;

plot(time, inf_data, 'b-', 'LineWidth', 1.5, 'DisplayName', 'Actual Data'); % Actual data

forecastTime = time(end) + (1:numForecasts); % Forecast time points

plot(forecastTime, forecastData, 'r-', 'LineWidth', 1.5, 'DisplayName', 'Forecasted Data'); % Forecasted data

xlabel('Time', 'FontSize', 14);

ylabel('Infected Cases', 'FontSize', 14);

title('Forecasting with VAR Model', 'FontSize', 16);

legend('Location', 'Best');

grid on;

hold off;

ERROR:

Error using varm/forecast (line 195)

The value of 'Y0' is invalid. Expected presample responses to be one of these types:

double

Error in plots (line 20)

YForecast = forecast(fitModel, numForecasts, 'Y0', presampleData);

Answer 1

[Pavl M.]

clc
clear all
close all

%% Step 1: Load data
data = xlsread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/1809848/DATA.xlsx');
time = double(data(1:400,1));          % Time column (1st column)
inf_data = double(data(1:400,2));      % Infected data column (2nd column)
d2 = double(data(1:400,3));
%% Step 2: Prepare data for VAR
% Since we are using univariate data, we treat it as a single-variable time series
% Create lagged data (you can modify the lag length)
lags = 2; % Number of lags
dataForVAR = [inf_data(1:end-lags);inf_data(lags+1:end)];
%% Step 3: Fit the VAR model
model = varm(1,lags);  % 2 lags, 1 sequence of variables
fitModel = estimate(model,dataForVAR); % Fit the model
%Identified model whose output is to be forecasted, specified as one of the following:
%Linear model — idpoly, idproc, idss, idtf, or idgrey
%Nonlinear model — idnlgrey, idnlhw, or idnlarx
%If a model is unavailable, estimate sys from PastData using commands 
% such as ar, arx, armax, nlarx, and ssest.
%% Step 4: Forecasting using the VAR model
summarize(fitModel)
numForecasts = 10;  % Number of steps to forecast
presampleData = dataForVAR(end-lags+1:end,:);  % The last lags data as presample
%opt = forecastOptions('InitialCondition','e'); % Add more options
%opt.InputOffset = 0;
YForecast = forecast(fitModel,presampleData,numForecasts); 
%or try using the estimated model and in-sample data as presample observations.
YForecast2 = forecast(fitModel,dataForVAR,numForecasts); 
% Rescale the forecasted data (if needed)
forecastData = YForecast(:, 1);  % First column is the forecast for the 'inf_data' series
forecastData2 = YForecast2(:,1);% Step 5: Plot the results
forecastTime = time(end)+(1:numForecasts);  % Forecast time points
%% Visualize anticipative foreseeing:
figure;
hold on;
plot(time,inf_data,'b-','LineWidth',1.5,'DisplayName','Actual Data');  % Actual data
plot(forecastTime,forecastData,'r-','LineWidth',1.5,'DisplayName','Forecasted Data');  % Forecasted data
xlabel('Time','FontSize',14);
ylabel('Infected Cases','FontSize',14);
title('Forecasting with VAR Model on presampled input','FontSize',16);
legend('Location','Best');
grid on;
hold off;
figure
plot(time,inf_data,'b-','LineWidth',1.5,'DisplayName','Actual Data');  % Actual data
plot(forecastTime,forecastData2,'r-','LineWidth',1.5,'DisplayName','Forecasted Data');  % Forecasted data
xlabel('Time','FontSize',14);
ylabel('Infected Cases','FontSize',14);
title('Forecasting with VAR Model on original input','FontSize',16);
legend('Location','Best');
grid on;
hold off;

[mallela ankamma rao]

Thank you for your response Pavl M sir. while i run the code i got the another error. i request please this problem sir

CODE:

% Clear workspace and setup

clear;

clc;

close all;

% % Step 1: Load data

data = xlsread('DATA.xlsx');

time = double(data(1:400, 1)); % Time column (1st column)

inf_data = double(data(1:400, 2)); % Infected data column (2nd column)

%% Step 2: Prepare data for VAR

% Since we are using univariate data, we treat it as a single-variable time series

% Create lagged data (you can modify the lag length)

lags = 2; % Number of lags

dataForVAR = [inf_data(1:end-lags),inf_data(lags+1:end)];

%% Step 3: Fit the VAR model

model = varm(1,2); % 2 lags, 1 variable

fitModel = estimate(model,dataForVAR); % Fit the model

%Identified model whose output is to be forecasted, specified as one of the following:

%Linear model — idpoly, idproc, idss, idtf, or idgrey

%Nonlinear model — idnlgrey, idnlhw, or idnlarx

%If a model is unavailable, estimate sys from PastData using commands

% such as ar, arx, armax, nlarx, and ssest.

%% Step 4: Forecasting using the VAR model

summarize(fitModel)

numForecasts = 10; % Number of steps to forecast

presampleData = dataForVAR(end-lags+1:end,:); % The last lags data as presample

%opt = forecastOptions('InitialCondition','e'); % Add more options

%opt.InputOffset = 0;

YForecast = forecast(fitModel,presampleData,numForecasts);

%or try using the estimated model and in-sample data as presample observations.

YForecast2 = forecast(fitModel,dataForVAR,numForecasts);

% Rescale the forecasted data (if needed)

forecastData = YForecast(:, 1); % First column is the forecast for the 'inf_data' series

forecastData2 = YForecast2(:,1);% Step 5: Plot the results

forecastTime = time(end)+(1:numForecasts); % Forecast time points

%% Visualize anticipative foreseeing:

figure;

hold on;

plot(time,inf_data,'b-','LineWidth',1.5,'DisplayName','Actual Data'); % Actual data

plot(forecastTime,forecastData,'r-','LineWidth',1.5,'DisplayName','Forecasted Data'); % Forecasted data

xlabel('Time','FontSize',14);

ylabel('Infected Cases','FontSize',14);

title('Forecasting with VAR Model on presampled input','FontSize',16);

legend('Location','Best');

grid on;

hold off;

figure

plot(time,inf_data,'b-','LineWidth',1.5,'DisplayName','Actual Data'); % Actual data

plot(forecastTime,forecastData2,'r-','LineWidth',1.5,'DisplayName','Forecasted Data'); % Forecasted data

xlabel('Time','FontSize',14);

ylabel('Infected Cases','FontSize',14);

title('Forecasting with VAR Model on original input','FontSize',16);

legend('Location','Best');

grid on;

hold off;

ERROR:

Error using varm/estimate (line 155)

The value of 'Y' is invalid. Expected Responses to be an array with number of columns equal to 1.

Error in plots (line 18)

fitModel = estimate(model,dataForVAR); % Fit the model

[Pavl M.]

OK

Yes, of course. It is what m for to contribute simple solutions to complex conundrums.

Today Fixed to (';' instead of your ',' for making data as a row, varm parameters swap to meet the standard how the function signature is defiend,forceast(...) params swap to meet the standard of how the function signature is defined in TCE):

...

dataForVAR = [inf_data(1:end-lags);inf_data(lags+1:end)];

%% Step 3: Fit the VAR model

model = varm(1,2); % 2 lags, 1 sequence of variables

fitModel = estimate(model,dataForVAR); % Fit the model

%Identified model whose output is to be forecasted, specified as one of the following:

%Linear model — idpoly, idproc, idss, idtf, or idgrey

%Nonlinear model — idnlgrey, idnlhw, or idnlarx

%If a model is unavailable, estimate sys from PastData using commands

% such as ar, arx, armax, nlarx, and ssest.

%% Step 4: Forecasting using the VAR model

summarize(fitModel)

numForecasts = 10; % Number of steps to forecast

presampleData = dataForVAR(end-lags+1:end,:); % The last lags data as presample

%opt = forecastOptions('InitialCondition','e'); % Add more options

%opt.InputOffset = 0;

YForecast = forecast(fitModel,presampleData,numForecasts);

...

*****

What do you get?

Next conundrum predicted:

What if you run predict to evaluate / backtest how predicted result matches the observed response of an estimated model:

[yp,x0,Predictor]= predict(fitModel,[crossvalidationdatum],numForecasts)

crossvalidationdatum = presampleData or dataForVAR or?

to check whether yp equals to YForecast to estimate your forecast accuracy?

What if the numForecasts growing / shrinking?

What if in forecast(...) you increase the presampleData from just 2 rows in 1 column orginally towards more rows and columns to get more evolutional funcional dependency (distribution) resemblings (subleties) of the original process/plant/environment model that produced such datum distribution?

*****

What if you exchange your varm model to

ar,(too simple model, losses in peculiarities)

arx, (too simple model, losses in peculiarities)

armax (not advisable to use, better use VARIMAX),

nlarx,

ssest (better also try this).

?

Then you can construct more sophisticated:

Mdl = msVAR(mc,mdl) %creates a Markov-switching dynamic regression model Mdl from:

dataForVAR1 = [inf_data(1:end-lags);inf_data(lags+1:end)];

dataForVAR2 = [d2(1:end-lags);d2(lags+1:end)];

lags = 4;

dataForVAR3 = [inf_data(1:end-lags);inf_data(lags+1:end)];

dataForVAR4 = [d2(1:end-lags);d2(lags+1:end)];

model1 = varm(1,2,'Covariance',cov1); % 2 lags, 1 sequence of variables

model2 = varm(1,4,'Covariance',cov2); % 4 lags

fitModel1 = estimate(model1,dataForVAR1); % Fit the model

fitModel2 = estimate(model1,dataForVAR2); % Fit the model

fitModel3 = estimate(model2,dataForVAR3); % Fit the model

fitModel4 = estimate(model2,dataForVAR4); % Fit the model

%Model with states:

mc = dtmc(dataForVARArray,'StateNames',["State1" "State2" "State3" "State4"])

mdl = [fitModel1; fitModel2;fitModel3;fitModel4];

Use msVAR to create a Markov-switching dynamic regression model from the switching mechanism mc and the state-specific submodels mdl.

Mdl = msVAR(mc,mdl) %Markov switching model with states

Or simply use sset instead of msVAR. Interesting research gap question is identified here can sset outperform msVAR in stated model accuracy ?

sys = ssest(datum,4); %if for 4 stated model estimation

Answer 2

[mallela ankamma rao]

Thank you for your respense sir. After modification, I run the code. Still I got the same error sir. I request you sir if possible please run the code and remove the error sir. Since two days, I tried sir but i did not get graph sir. This graph is necesssary for research article sir.

ERROR:

Error using varm/estimate (line 155)

The value of 'Y' is invalid. Expected Responses to be an array with number of columns equal to 1.

Error in plots (line 144)

fitModel = estimate(model, dataForVAR'); % Fit the model (data needs to be transposed)

CODE:

home

clear all

close all

%% Step 1: Load data

data = xlsread('DATA.xlsx');

time = double(data(1:400,1)); % Time column (1st column)

inf_data = double(data(1:400,2)); % Infected data column (2nd column)

d2 = double(data(1:400,3)); % Additional data if necessary

%% Step 2: Prepare data for VAR

lags = 2; % Number of lags

dataForVAR = [inf_data(1:end-lags); inf_data(lags+1:end)];

%% Step 3: Fit the VAR model

model = varm(1, lags); % 1 sequence of variables, 2 lags

fitModel = estimate(model, dataForVAR'); % Fit the model (data needs to be transposed)

%% Step 4: Forecasting using the VAR model

summarize(fitModel)

numForecasts = 100; % Number of steps to forecast

% Prepare the presample data (last lags data)

presampleData = dataForVAR(end-lags+1:end, :); % Matrix format for VAR forecast

% Forecast for the next 10 time steps

YForecast = forecast(fitModel, numForecasts, presampleData);

% Rescale the forecasted data (if needed)

forecastData = YForecast(:, 1); % First column is the forecast for the 'inf_data' series

%% Step 5: Plot the results

forecastTime = time(end) + (1:numForecasts); % Forecast time points

% Create a new figure for plotting

figure;

hold on;

plot(time, inf_data, 'b-', 'LineWidth', 1.5, 'DisplayName', 'Actual Data'); % Actual data

plot(forecastTime, forecastData, 'r-', 'LineWidth', 1.5, 'DisplayName', 'Forecasted Data'); % Forecasted data

xlabel('Time', 'FontSize', 14);

ylabel('Infected Cases', 'FontSize', 14);

title('Forecasting with VAR Model', 'FontSize', 16);

legend('Location', 'Best');

grid on;

hold off;

[Pavl M.]

Why, for what You invoked

estimate(model, dataForVAR'); with dataForVAR transposed (' sign)?

while at the time of your latest comment I've already updated(corrected) the my answer:

estimate(model, dataForVAR);

after I had constructed the correct column vector:

dataForVAR = [inf_data(1:end-lags);inf_data(lags+1:end)];

which is a 1 column vector and estimate "eats" 1 column vector.

With no transposed version, run it what you get?

I think you used un-updated versions.

I corrected it initially:

That is your dataForVAR in green.

Are you sure about it. It contains 1 peak, gaussian-like distribution and seasonable(periodic) components.

What is your time scale/units? (start date, end date?)

I think the data to modify to

Human operator can analyse and predict, the best forecast by natural intelligence on this data is that it will further grow up(uptrend) till some second following peak then will go downtrend.

The dataForVAR so far is more like for evaluation of how varm(...) estimated the underlying model and for forecast m on it ongoing construction of the data...

Let me kow which more help is required? It should be OK now and in the future.

Please run it and let me know what you get, there are a few very valued R&D workstreams to proceed and continue furthermore.

[mallela ankamma rao]

Thank you for your response sir. Yes sir the graph is very good. I want that same graph sir. But when irun the code i got the error sir. My matlab version is R2021b.

CODE:

home

clear all

close all

%% Step 1: Load data

data = xlsread('DATA.xlsx');

Dlength = 400;

time = double(data(1:Dlength,1)); % Time column (1st column)

inf_data = double(data(1:Dlength,2)); % Infected data column (2nd column)

d2 = double(data(1:Dlength,3));

%% Step 2: Prepare data for VAR

% Since we are using univariate data, we treat it as a single-variable time series

% Create lagged data (you can modify the lag length)

lags = 2; % Number of lags

dataForVAR = [inf_data(1:end-lags);inf_data(lags+1:end)];

%% Step 3: Fit the VAR model

model = varm(1,lags); % 2 lags, 1 sequence of variables

fitModel = estimate(model,dataForVAR); % Fit the model

%Identified model whose output is to be forecasted, specified as one of the following:

%Linear model — idpoly, idproc, idss, idtf, or idgrey

%Nonlinear model — idnlgrey, idnlhw, or idnlarx

%If a model is unavailable, estimate sys from PastData using commands

% such as nlarx, and ssest.

%% Step 4: Forecasting using the VAR model

summarize(fitModel)

numForecasts = 10; % Number of steps to forecast

presampleData = dataForVAR(end-lags+1:end,:); % The last lags data as presample

dataForVAR2 = inf_data(1:end);

%opt = forecastOptions('InitialCondition','e'); % Add more options

%opt.InputOffset = 0;

YForecast = forecast(fitModel,presampleData,numForecasts);

%or try using the estimated model and in-sample data as presample observations.

YForecast2 = forecast(fitModel,dataForVAR,numForecasts);

YForecast3 = forecast(fitModel,dataForVAR2,numForecasts);

% Rescale the forecasted data (if needed)

forecastData = YForecast(:, 1); % First column is the forecast for the 'inf_data' series

forecastData2 = YForecast2(:,1);

forecastData3 = YForecast(:,2);

forecastData4 = YForecast2(:,2);

forecastData5 = YForecast3(:,1);

forecastTime = time(end)+(1:numForecasts); % Forecast time points

% Step 5: Plot the results

%% Visualize anticipative foreseeing:

figure;

hold on;

plot(time,inf_data,'b-','LineWidth',1.5,'DisplayName','Actual Data'); % Actual data

plot(forecastTime,forecastData,'r-','LineWidth',1.5,'DisplayName','Forecasted Data'); % Forecasted data

xlabel('Time','FontSize',14);

ylabel('Infected Cases','FontSize',14);

title('Forecasting with VAR Model on presampled input','FontSize',16);

legend('Location','Best');

grid on;

hold off;

figure

plot(inf_data,'b-','LineWidth',1.5,'DisplayName','Actual Data'); % Actual data

plot(forecastData5,'r-','LineWidth',1.5,'DisplayName','Forecasted Data'); % Forecasted data

xlabel('Time','FontSize',14);

ylabel('Infected Cases','FontSize',14);

title('Forecasting with VAR Model on original input','FontSize',16);

legend('Location','Best');

grid on;

hold off;

ERROR:

Error using varm/forecast (line 195)

The value of 'numperiods' is invalid. Expected forecast horizon to be a scalar.

Error in plots (line 125)

YForecast = forecast(fitModel,presampleData,numForecasts);

Answer 3

[Pavl M.]

It's still OK, quite smooth and not problematic to solve complete and resolve in full.

There are 2 different forecast functions with the same name:

https://www.mathworks.com/help/ident/ref/idmodel.forecast.html

[yf,x0,sysf,yf_sd,x,x_sd] = forecast(sys,PastData,K)

and

https://www.mathworks.com/help/econ/arima.forecast.html

[Y,YMSE,V] = forecast(Mdl,numperiods,Y0)

where Mdl is the system and Y0 is your presample.

One in Econometrics toolbox and 1 in system identification toolbox. In econometrics they assume by default ARIMA and GARCH composite model.

System identification is better, since it covers more real world cases(more general). While forecast in Econometrics mainly fitted for ARIMAX model. That 's why even before I putted in my answer the hint to try other estimators of model, in addition to varm.

I know why your system invoked the forecast from Econometrics. That is mechanism of function dispatch in TCE, identified it due to the input Mdl of type varm.

I assumed the first one as most reasonable per se.

It's still ok, the solution is just swap the tail 2 function arguments places.

Before R2023 version they may be calling the Econometrics 1 by default

Check and let me know.

Which toolboxes do you have which you will be needing?

Swap presampleData and numForecasts places in it.

What on TCE R2024b?

If on R2021b value of 'numperiods' Expected forecast horizon to be a scalar.

(line 121)

YForecast = forecast(fitModel,presampleData,numForecasts);

Try to swap parameters

numForecasts is the predictionHorizon, which is scalar

numForecasts = 10;

I think I esablished robust trust. Let's build solid understanding. Will you be needing it in R and Python as well?

Turst me.

I'll be happy to help, to debug your code.

I think now you have succeeded to launch it and get intial results.

What do you get now?

[mallela ankamma rao]

Good evening Pavl M sir. Thank you verymuch for your patience and kindness sir. I run the code sir. The previous error resolved sir but new error came sir. Please resolve this also sir.

CODE:

home

clear all

close all

%% Step 1: Load data

data = xlsread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/1809848/DATA.xlsx');

Dlength = 400;

time = double(data(1:Dlength,1)); % Time column (1st column)

inf_data = double(data(1:Dlength,2)); % Infected data column (2nd column)

d2 = double(data(1:Dlength,3));

% areTimestampsRegular = isregular(data,"days")

% areTimestampsSorted = issorted(data)

% w = convert2weekly(data,Aggregation="mean")

% areTimestampsWRegular = isregular(w,"weeks")

%% Step 2: Prepare data for VAR

% Since we are using univariate data, we treat it as a single-variable time series

% Create lagged data (you can modify the lag length)

lags = 2; % Number of lags

dataForVAR = [inf_data(1:end-lags);inf_data(lags+1:end)];

%% Step 3: Fit the VAR model

model = varm(1,lags); % 2 lags, 1 sequence of variables

fitModel = estimate(model,dataForVAR); % Fit the model

%Identified model whose output is to be forecasted, specified as one of the following:

%Linear model — idpoly, idproc, idss, idtf, or idgrey

%Nonlinear model — idnlgrey, idnlhw, or idnlarx

%If a model is unavailable, estimate sys from PastData using commands

% such as nlarx, and ssest.

%% Step 4: Forecasting using the VAR model

summarize(fitModel)

numForecasts = 100; % Number of steps to forecast

% numpreobs = fitModel.P % Required presample length as per ARIMAX modeling

% w0 = w(1:numpreobs,:); % Estimation presample

% w1 = w((numpreobs+1):(end-numForecasts),:);

% % In-sample for estimation and presample for forecasting

% w2 = w((end-numperiods+1):end,:); % Holdout sample

presampleData = dataForVAR(end-lags+1:end,:); % The last lags data as presample

dataForVAR2 = inf_data(1:end);

fitModel2 = estimate(varm(1,0),dataForVAR); % Fit the model

%opt = forecastOptions('InitialCondition','e'); % Add more options

%opt.InputOffset = 0;

YForecast = forecast(fitModel,numForecasts,presampleData);

%or try using the estimated model and in-sample data as presample observations.

YForecast2 = forecast(fitModel,numForecasts,dataForVAR);

YForecast3 = forecast(fitModel,numForecasts,dataForVAR2);

% Rescale the forecasted data (if needed)

forecastData = YForecast(:, 1); % First column is the forecast for the 'inf_data' series

forecastData2 = YForecast2(:,1);

forecastData3 = YForecast(:,2);

forecastData4 = YForecast2(:,2);

forecastData5 = YForecast3(:,1);

forecastTime = time(end)+(1:numForecasts); % Forecast time points

% Step 5: Plot the results

%% Visualize anticipative foreseeing:

figure;

hold on;

plot(time,inf_data,'b-','LineWidth',1.5,'DisplayName','Actual Data'); % Actual data

plot(forecastTime,forecastData,'r-','LineWidth',1.5,'DisplayName','Forecasted Data'); % Forecasted data

xlabel('Time','FontSize',14);

ylabel('Infected Cases','FontSize',14);

title('Forecasting with VAR Model on presampled input','FontSize',16);

legend('Location','Best');

grid on;

hold off;

figure

plot(inf_data,'b-','LineWidth',1.5,'DisplayName','Actual Data'); % Actual data

plot(forecastData5,'r-','LineWidth',1.5,'DisplayName','Forecasted Data'); % Forecasted data

xlabel('Time','FontSize',14);

ylabel('Infected Cases','FontSize',14);

title('Forecasting with VAR Model on original input','FontSize',16);

legend('Location','Best');

grid on;

hold off;

ERROR:

Index in position 2 exceeds array bounds (must not exceed 1).

Error in plots (line 143)

forecastData3 = YForecast(:,2)

[Pavl M.]

Dear Friend,

Yes, OK:

% Rescale the forecasted data (if needed)
forecastData = YForecast(:, 1);  % First column is the forecast for the 'inf_data' series
forecastData2 = YForecast2(:,1);
%forecastData3 = YForecast(:,2);
%forecastData4 = YForecast2(:,2);
forecastData5 = YForecast3(:,1);
forecastTime = time(end)+(1:numForecasts);  % Forecast time points

% Step 5: Plot the results
%% Visualize anticipative foreseeing:
figure;
hold on;
plot(time,inf_data,'b-','LineWidth',1.5,'DisplayName','Actual Data');  % Actual data
plot(forecastTime,forecastData,'r-','LineWidth',1.5,'DisplayName','Forecasted Data');  % Forecasted data
xlabel('Time','FontSize',14);
ylabel('Infected Cases','FontSize',14);
title('Forecasting with VAR Model on presampled input','FontSize',16);
legend('Location','Best');
grid on;
hold off;

figure
plot(inf_data,'b-','LineWidth',1.5,'DisplayName','Actual Data');  % Actual data
plot(forecastData5,'r-','LineWidth',1.5,'DisplayName','Forecasted Data');  % Forecasted data
xlabel('Time','FontSize',14);
ylabel('Infected Cases','FontSize',14);
title('Forecasting with VAR Model on original input','FontSize',16);
legend('Location','Best');
grid on;
hold off;

What, which to complete?

[mallela ankamma rao]

Thank for your response sir. I tried this also sir but i got these figures sir which are entirerly to different to your figures sir.

[Pavl M.]

So you got the desired figures(plots).

Please understand. I build solid understanding.

1) Let me know more details about the Project, Research Agenda you need, the research Titles, topics, scope, domain, importance, impact and desired outcomes/service/product you need to get.

2)The green figures I plotted are just for the input data to forecaster. The figures that you produced with our mutual effort with my factual improvements to you ongoing R&D are figures after the estimator and forecaster start working upon my constructive adjustments(It's okay, please contact me more in private if there is a need to manuscript, research work and paper collaboration, cooperation, co-authorship on reciporcally convenient terms: it is very beneficial will be to you if you take me on board: 2ce the heads, hands and machines = *2 more quality in output). The both my and your plots are OK, your just plot is after the model estimation and forecast + see I added there in DatumModelling2.m more features to add to the Project + I am in progress of more understanding and r&d SciTech value to add to it.

I can make for you more on this specific professional expertise project:

a) Find and open 2D or 3D data, make internal system model estimation using built-in nlarx,or if you want to really enhance it to catch more determinism by forecasting on a blend of a few signals, like input-output, cause-effect sys = ssest(u,y,nx,'Ts',Ts) (very promissing 2, much better than ar, arx, arima, sarima,). With nx = number of states, u - input, y - output

b) With estimated model make forecast using the second System Identification toolbox function,

c) Make Grange, novelties tests on estimates,

d) Check forecast accuracy with predict, compare, sim built-in functions, make crossvalidations, evaluate and compare model estimations and forecasting powers,

e) Figure out what is the gist and purpose of lags in the process,

f) Add more informative features additions, like RM, MAE, MSE, RMSE, kFoldLoss to the base process and make more interactive plots.

It may be better if we switch to more bening commerical/industrial mode and join/merge efforts towards better Control Systems Engineering specific projects and assingments for modest remuneration as a salary (just fixed monthly rate) grant, stipend, necessary financial aid I request,AUTOSAR, NVIDIA RTX4070 or bettr, DLSS3 or better domain understanding and making for sale. There is also recently I found a few SOA research directions in Deep Learning. Let me know what are the priorities, which critical help and which complete parts will be most necessary?

I have a few more large novel codes in Python for similar prediction on multi-dimensional data with boosted regressions, can give you if you/around you there in your the well-enclosed professional circle will require/be interested/be needing it.

Kindly accept my answer at least here firstly and than if you need graduated engineer researcher on aforementioned terms, contact me more in private.

[mallela ankamma rao]

Thank you very much Pavl M sir. Please send your mail or whatsupp number sir. I will contact you and explain about my research paper sir.

[Pavl M.]

Contact me more for work at: [e-mail], my genuine ms skype instead of whatsapp: join.skype.com/invite/oXnJhbgys7oW ™®©

@mallela ankamma rao,

Please carefully think about the honest real future business roadmap:

Due to we met for specific work here, no hacking please, only ethically correct paid employment.

Have you proof-read well the recent PuMed article published at 27 September 2024?

Have you published also at Springer Nature ™®© ?

Who will be glad to hire on proof-reading services?

How SAIUGR model is accurate, what else high value models to help with?

Do your still need refinement and more crisp adjustments to SIDAREV model?

How much for more sophisticated and utile visualization and processing within the scope?

Can we uphold and maintain focus group with Chilean, Portugeese and Spanish investors/stakeholders, which components, visual artefacts, digital items they will be happy to pay for under the free trade framework proposed?

Which specific commitment who will be glad to transact salary via onlne banking or via WU?

Which team is there to publish at PubMed?

What if to improve Λ Human recruitment rate?

What on Optimal Control who will be needing most valuable?

I can do in Technical Computing Environment, Numerical Computing, Multi-Paradigm Programming,

tending to holistic dynamics and statics analysis, which ICER, NFSI, other metrics to implement?

Which objective functions to care for?

What on me to do and look what I did before in order to get more necessary substantial supplies?