Error using load Number of columns on line 2 of ASCII file

Hello everybody !

I want to make a prediction by using Extreme learning Machine (ELM), but i encountered a problem when reading data from my file excel. My data are in the following form:

Column 1: contains the date format (dd / mm / yyyy hh: mm)

Column 2: contains numeric values.

Here is the code :

train_data=load(TrainingData_File);
T=train_data(:,1)';
P=train_data(:,2:size(train_data,2))';
clear train_data;

Here is the error :

Error using load
Number of columns on line 2 of ASCII file SolarData_training.xlsx must be the same as previous lines.

I use Matlab R2014 a.

Thank you so much !

Answer 1

[Star Strider]

It is best to not use the load function with Excel files. Use xlsread to import your data. Use the datenum function to convert your date and time strings to date numbers and other formats.

Answer 2

[Niels]

matlab tries to create a matrix with 2 rows. obviously your second row has more or less colums than the first row, so your matrix looks like

A =
     1     0     2
     3     4

such a matrix does not exist, thats why an error returns

use xlsread and the missing colums will be filled with NaNs

Answer 3

[Walter Roberson]

I would suggest readtable() for this purpose. As you are using R2014a, the times will not be converted to datetime objects; they will either show up as excel date numbers or as strings, depending on your operating system.

If you end up with an excel date number, then if you have the Finance toolbox you can use https://www.mathworks.com/help/finance/x2mdate.html

If you do not have that toolbox, then you can add datenum() of Dec 31 1899 to the date. But you have to be careful; see https://www.mathworks.com/matlabcentral/answers/254-converting-excel-serial-number-of-time-to-hh-mm-ss-format#answer_353

Answer 4

[Hamza Ali]

Thanks very much for your reaction, the problem is solved using the xlsread function. However, I get only one column that the numeric values, but the one containing the date is unrecoverable.

Best regards.

[Star Strider]

The date information is there. You have to request at least two outputs, the numeric data (the first output) and the ‘string’ output (the second output).

The second (‘string’) output contains the date strings as a cell array of strings. You can use the cell array as input to datenum to convert them to date numbers, and from there, convert them to other date formats if you want to.

See the documentation for xlsread and datenum for details.

[per isakson]

"However, I get only one column that the numeric values, but the one containing the date is unrecoverable." &nbsp I guess, your problem is related to the difference between the default reference dates of Excel and Matlab.

See Import and Export Dates to Excel Files

MATLAB serial date number     January 0, 0000
Excel for Windows             January 1, 1900
Excel for the Macintosh       January 2, 1904

[Walter Roberson]

It is worse than that, Excel for Windows thinks Feb 29 1900 was a leap year but it was not, so the adjustment depends on whether your dates refer to dates after Feb 28 1900.

Answer 5

[Hamza Ali]

Hello everyone !

I followed your advice literally, which led me to the following result:

   %%%%%%%%%%%Load training dataset
    [x,y]=xlsread(TrainingData_File);
    y=datenum(y);
    T = x';
    P = y';
    %%%%%%%%%%%Load testing dataset
    [z,t]=xlsread(TestingData_File);
    t = datenum(t);
    TV.T=z';
    TV.P=t';

Fortunately, the function is executed successfully. But are the changes on the code of the function did not impact the results obtained ? because I get a result that appears twice as follows:

TrainingTime =

    2.3088
TrainingAccuracy =

   57.2277
TestingTime =

    0.2028
TestingAccuracy =

   57.0400
TrainingTime =

    2.3088
TestingTime =

    0.2028
TrainingAccuracy =

   57.2277
TestingAccuracy =

   57.0400

Sincerely, thanks to your answers, I advanced in my works and I saved a lot of time. Thank you very much.