Increasing Speed of Execution

The code below analyzes the slope of a dataset. We know that the graph of this dataset has a very distinct point where the slope begins rising much faster. To identify this point in MATLAB, we have written the follwoing code...

Deviation = zeros(length(Temperature), 1);
for i = 1:length(Temperature)
    Temperature_Avg = sum(Temperature(1:i))/i;
    Deviation(i) = (Temperature(i) - Temperature_Avg)^2;
end

Avg_Slope = zeros(length(Deviation)/5, 1);
for j = 1:5:length(Deviation)
    Avg_Slope(j + 4) = (sum(Deviation(j:1:(j + 4)))) / 5;
end

n = 1;
Partial_Avg_Slope = Avg_Slope(5) / n;
Run_Avg_Slope = sum(Avg_Slope(1:1:(5 * n))) / n;
while Partial_Avg_Slope <= (4 * Run_Avg_Slope)
    n = n + 1;
    Run_Avg_Slope = sum(Avg_Slope(1:1:(5 * n))) / n;
    Partial_Avg_Slope = Avg_Slope(5 * n);
end
t_s = Time((n * 5), 1);

if Temperature(end) > Temperature(1) % Parameter for heating system
    y_L = sum(Temperature(1:1:(n * 5)))/(n * 5);
elseif Temperature(end) < Temperature(1) % Parameter for cooling system
    y_H = sum(Temperature(1:1:(n * 5)))/(n * 5);
end

The code achieves it's goal but it is very slow, especially since it has to be executed for 100 datasets. Most of that run time comes from the line that caluclates Temperature_Avg. Is there any way that that line can be written so that it runs faster? Temperature and Time are row vectors with 10240 entries.

Answer 1

[Steven Lord]

Try the ischange function.

[Jacob Rensner]

Thank you. I think that is exaclty the type of function we were needing.