Times-C

Times-C is a GPU-accelerated k-Shape time series clustering algorithm.

Installation

CUDA toolkit

To install CUDA toolkit please use this link.

Complier

It is recommended to use a g++ compiler that supports C++14 or above.

Dataset

Datasets can be obtained from the following websites:

Welcome to the UCR Time Series Classification/Clustering Page

Time Series Classification Website

k-Shape & tslearn algorithm

k-Shape algorithm can be obtained from the following websites:

[https://github.com/TheDatumOrg/kshape-python]

tslearn algorithm can be obtained from the following websites:

[https://github.com/tslearn-team/tslearn]

Times-C GPU

Compilation

Compile the code using the following command:

cd Code
make

Execution

Run the code using the following command:

cd bin

./test -d devNum /path/to/dataset k flag init_chose

Description of the arguments:

• -d devnum: Specify the device the code runs on
• /path/to/dataset : The relative or absolute path to the file containing time series
• k : Number of clusters
• flag: Whether the first column of the specified dataset is a label. If the first column is not a label, set it to 1, otherwise set it to 0. The first column of UCR database is a label, while the first column of UCR and UAE databases is not a label
• init_chose: When the value is 0, it represents random initialization of cluster indices. When the value is 1, it represents random initialization of cluster centers.
• Example: ./test -d 0 ../data/Part_of_HandOutlines 2 0 0

Times-C CPU

Execution

cd Code/test

python test_timesC.py /path/to/dataset k flag 

Description of the arguments:

• /path/to/dataset : The relative or absolute path to the file containing time series(Suggest using absolute paths)
• k : Number of clusters
• flag: Whether the first column of the specified dataset is a label. If the first column is not a label, set it to 1, otherwise set it to 0. The first column of UCR database is a label, while the first column of UCR and UAE databases is not a label