how to remove outliers in large data sets?

I am unable to open example code of outliers (openExample('matlab/RemoveOutliersInVectorExample') ) and openExample('matlab/DetermineOutliersWithStandardDeviationExample') also.

I had large datasets of power load for three years at 30min interval, I want to remove the outliers poitns which is affecting my forecasting error.

any help to remove the outliers in such datasets would be appreciated.

Thanks

Reference image is attached which shows the outliers datasets in upper side of image, reference to these points the error is also high (as lower side of image).

Thanks again

[Image Analyst]

How large is the data set? How many gigabytes? Can you attach a smaller set (less than 5 MB) in a .zip file?

[MUKESH KUMAR]

Not in GB but its five years data having outliers data in patters, i am attaching data file here

Answer 1

[Image Analyst]

Try this:

data = readmatrix('Copy of data.xlsx');
x = data(:, 1);
y = data(:, 2);
% Plot just the first cycle.
last = round(70000/3) 
x = x(1:last);
y = y(1:last);
subplot(2, 1, 1);
plot(x, y, 'b-')
grid on;
title('Showing One Cycle Only')
% Smooth the data.
windowWidth = 2001;	% Some large odd number.
smoothY = movmean(y, windowWidth);
hold on;
plot(x, smoothY, 'r-', 'LineWidth', 3)
% Compute difference between actual and smoothed.
diffy = y - smoothY;
subplot(2, 1, 2);
plot(x, diffy, 'b-');
grid on;
% Detect outliers as having a MAD of more than 900
outlierIndexes = abs(diffy) > 900;
% Plot outliers as red dots over the original data.
subplot(2, 1, 1);
hold on
plot(x(outlierIndexes), y(outlierIndexes), 'r.', 'MarkerSize', 7);

% Now remove outliers from x and y
x(outlierIndexes) = [];
y(outlierIndexes) = [];

[MUKESH KUMAR]

This is very helpful and I understand the code and applied in my problem

I also tried to fill outliers data like openExample('matlab/DetectOutliersWithSlidingWindowRmoutliersExample')

but the cmd is not working in 2018a version, its applicable on 2018b version.

''rmoutliers" cmd not working

Is there any way to how to fill these outliers with using movmean or movmedian?

Thanks

[Image Analyst]

You must have a version so old that rmoutliers was not in it yet. However you can do it manually. Just smooth the curve and subtract it from your data and threshold like I did. I didn't use rmoutliers.