comparing time series data and grouping

I have radiation data in time series in every 5 min during month so I wanted to compire days with each other (compire time serieses) and group them according to similerity.

Could you kinldy help me with how to do this annalysis in matlab. I am new to matlab

I need;

1. Compire differant time seriesies of differant dates and group the similerity, somthing like this;

Day Grouping

day01 a

day02 ab

day 03 b

..... .....

2. Need to find the time seriese with higerst values and low variability

3. Need to find the time seriese with lowest values and low variabilty

Data stored in a mat file DATA (data are nonlinear)

(:,1) is date (:,2) is time and (:,3) is the reading i have. (readings are changing with time)

Date Time Readings

01 06.00 0.500

01 06.05 0.450

01 ... ...

01 19.00 0.235

02 06.00 0.700

02 06.05 0.650

02 ... ...

02 19.00 0.335

... ... ...

31 19.00 0.432

Could you kindly helpme find the test ishould use to do this and matlab codes to do this. I am Highly appriciating your advicers ..

Answer 1

[Arjun]

Hi @Amila,

I understand that you have a dataset containing radiation data and you want to group similar data and perform some analysis on top of it.

Since nothing is mentioned about the algorithm for grouping data, lets assume “k-means” clustering as the grouping algorithm. We can divide the entire task into three main chunks which are as follows:

• Loading and preparing data.
• Comparing and grouping time series by similarity.
• Identifying time series with highest and lowest values with low variability.

Kindly refer to the code below for better understanding:

% Load the data, part 1 of the task 

load('DATA.mat');  

dates = DATA(:, 1); 

times = DATA(:, 2); 

readings = DATA(:, 3); 

dataTable = table(dates, times, readings); 

 

% Get unique days 

uniqueDays = unique(dates); 

 

% Reshape data into matrix form where each row is a day 

numDays = length(uniqueDays); 

numTimePoints = sum(dates == uniqueDays(1));  

dataMatrix = reshape(readings, numTimePoints, numDays)'; 

 

% Normalize data (optional, depending on your data characteristics) 

dataMatrix = zscore(dataMatrix, 0, 2); 

 

% Perform k-means clustering- grouping the data, part 2 of task 

numClusters = 3;  

[idx, ~] = kmeans(dataMatrix, numClusters); 

 

% Create a grouping table 

groupingTable = table(uniqueDays, idx, 'VariableNames', {'Day', 'Group'}); 

disp(groupingTable); 

 

% Calculate mean and standard deviation for each day- part 3 of the task 

means = mean(dataMatrix, 2); 

stdDevs = std(dataMatrix, 0, 2); 

 

% Define a threshold for low variability 

threshold = 0.1;  

lowVarIndices = stdDevs < threshold; 

if any(lowVarIndices) 

    highMeanLowVar = uniqueDays(lowVarIndices & means == max(means(lowVarIndices))); 

    lowMeanLowVar = uniqueDays(lowVarIndices & means == min(means(lowVarIndices))); 

else 

    highMeanLowVar = []; 

    lowMeanLowVar = []; 

end 

 

disp('Day with highest mean and low variability:'); 

disp(highMeanLowVar); 

 

disp('Day with lowest mean and low variability:'); 

disp(lowMeanLowVar); 

Kindly refer to the following documentation link for more information on “k-means” algorithm: https://www.mathworks.com/help/stats/kmeans.html

I hope this will help!