Convert date from number to format in matllab

I have original time and date format as mentioned below:

2/17/2020 9:48:18 AM

I did forecasting using this , and after forecasting output date in the plot appears in number format like :

43878.4085416667

How can i convert it back to original date and time format ?

[Walter Roberson]

How did you do the forecasting with it?

Did you convert the date/time t numeric using datenum() ? If so, only do that if the forecasting routines do not support datetime objects, and otherwise use datetime() .

[NN]

yes, i converted using datenum()

Data=readtable('data.xlsx'); 
testdates = Data{1:end,1};
date_num = datenum(testdates);
date_num2 = datenum(testdates)+table2array(Data(1:end,2))./(24*3600);

[NN]

then i used below code to plot it :

figure
plot(out)
hold on
plot(in,'.-')
hold off

[NN]

but it appears like serial numbers :i want to convert it to date and time format :i tried the below code but unable to plot since it is character array(pic attached) :If i convert it to double again it becomes NAN since it has string values:

o=datestr(in);
S = double(out);
S1=datestr(S);

figure
plot(out)
hold on
plot(in,'.-')
hold off

[Walter Roberson]

date_num2 = testdates + seconds(Data{:,2});

Now if possible use date_num2 in the forecasting, and plot with

plot(date_num2, in)

[NN]

thank you for your response.It worked for input date ,but didnt work for the output

out is date value generated after forecasting , and how can i convert this output back to date time format .

[Walter Roberson]

How are you generating the date value from forecasting ?

The values near 43878.4085416667 suggest to me,

datetime(43878.4085416667, 'convertfrom', 'excel')

ans = datetime
   17-Feb-2020 09:48:18

but why are you getting Excel dates at all???

[NN]

This is the output i am getting . i need to convert it back to date time format

[NN]

i tried converting it to character array using below code ,

S = double(out);

S1=datestr(S);

but still not able to plot

Answer 1

[Walter Roberson]

outd = datetime(out, 'convertfrom', 'excel');
plot(outd)

[NN]

thank you .It worked

[Walter Roberson]

Your code should be changed so that it does not produce Excel date numbers from the forecast. However, you are not being cooperative in identifying how it is getting Excel date numbers as output values, so I stopped caring.

[NN]

:-(

I will definitly change the code .But please advise how to do it .i am not an expert in programming , thats why

[Walter Roberson]

The first step would be to describe how you are currently doing the forecasting. Which I already asked about twice before.

[NN]

LSTM forecasting

[Walter Roberson]

The error is on line 7, character #11.

[NN]

sorry , i didnt understand

[Walter Roberson]

This is the information we, as outside observers, have so far:

1. You have a datetime() vector named in
2. Your results from your forecasting is named out
3. Your results are in Excel date numbers
4. You are using lstm forecasting.

(4) tells us that you must have created layers, and called trainNetwork(), and probably called predictAndUpdateState() https://www.mathworks.com/help/deeplearning/ug/time-series-forecasting-using-deep-learning.html

However, it does not tell us what you passed to trainNetwork(), and it does not tell us what layers you used.

So... we need to read your mind to figure that out. And when I read your mind, and your memory of what code you wrote, I figured out that your bug is in character #11 of line 7 of your code.

[NN]

so sorry ,

i tried to do LSTM in the same way as in https://www.mathworks.com/help/deeplearning/ug/time-series-forecasting-using-deep-learning.html

[Walter Roberson]

I was right, the bug is in character #11 of line 7 of your code. (Prove me wrong!)