Cluster of precipitation time series

9 views (last 30 days)
CJ
CJ on 11 Sep 2019
Commented: Mahesh Taparia on 20 Sep 2019
Hi guys,
I have a precipitation time series of several years and I would like to cluster the information automatically by seasons (Autumn, Winter, Spring, Summer). Using the kmeans algorithm (kmeans(X,4)), I only get the whole domain of the data (i.e. all the years) split in 4 equally spaced subdomains. However, I would like to split each year contained in the whole dataset in the four seasons.
Any ideas from your side?
R
  5 Comments
CJ
CJ on 12 Sep 2019
It is an array of two columns. The first column is time and the second one is precipitation (mm of rain), which is cyclic in time due to the year seasonality. Both variables are double type.
darova
darova on 12 Sep 2019
Can you show your data and data you want to see?

Sign in to comment.

Answers (2)

Mahesh Taparia
Mahesh Taparia on 17 Sep 2019
Hi CJ,
Since you want to cluster the precipitation data into 4 seasons in year wise format, you should take year wise data and then perform the K-means clustering with 4 classes for each season. Save the clustered data year wise. Hope this helps.

CJ
CJ on 20 Sep 2019
Hi Mahesh,
Thanks for your input, but I don't follow it. If we split each season in four classes and we have four seasons we ultimately result in 16 classes which is not my purpose.
R
  1 Comment
Mahesh Taparia
Mahesh Taparia on 20 Sep 2019
Hi,
As per my answer, you will have 4 classes only. Follow these steps:
  1. Take the precipitation data of a year, cluster it into 4 classes as per season.
  2. Take the second year data and again cluster it into 4 classes.
  3. Apppend these 2 years of data of same seasons into 4 classes with the same cluster head. It will result in 4 classes only.
  4. Repeat the steps upto n'th year data.
  5. Finally, you will have 4 classes only.

Sign in to comment.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!