Motifs discovery into spatial time series

Amin Bazaz, Heraldo Borges, Eduardo Ogasawara

2018-08-06

This package STSMotifs allows to perform a research of motif into spatial time series. The main purpose is to find a way to handle the issue of large amouts of data. The package offers a way to do this research quickly and efficiently. To find the motifs, the Combined Series Approach (CSA) is used. The process is decomposed by several steps :

  1. generation of candidates
  2. Check and filter the candidates by some tresholds
  3. rank the selected candidates

Input

To use functions of this package, some inputs are needed. The quality of outputs depends strongly by these parameters, the user has to pay attention.

The dataset contain only numerics values. Columns represent the space and rows the time.

#>        1    2    3     4     5     6    7     8    9   10
#> 360  737 1350  869   750  1138   758 1006  1095   99  -83
#> 361  283  565  504   317  1849   944  -80  -895 -936  906
#> 362 -118 -375 -564  -803   870   472 -922 -1009 -698  741
#> 363 -696 -844 -654 -1303  -474  -591 -262  1034 1012  376
#> 364 -251 -622  -14  -587 -1108 -1401  404  1545 1696  247
#> 365  645  -10   -4   411  -858 -1261 -574  -329 -367 -680

A part of the process is applied into blocks (subset of the original dataset). With the tslice (“Time slice” number of rows in each block) and sslice (“Space slice” number of columns in each block), the user can specify a block shape (vertical rectangle, square or horizontal rectangle).

Kappa (Spacial Occurence) represents the minimal number of motif occurences in the different columns. Sigma (Global Occurence) represents the minimal number of motif occurences in the throughout the dataset.

Description of each step

Generation of candidates

This step create blocks. After that, all this process is adapted in a single time series. Thereby each columns of the block are combined to create a single big time series. The output of this step are information about each blocks :

Treatment of candidates

This step requires the candidates and the kappa and sigma thresholds. All information about motifs are extracted and manipulated At the end, the motifs that passed the restriction of the 2 thresholds are stored into a list of motif. Each pattern has as information :

Once these motifs created, a filtering is applied.

Rank the selected patterns

The previous step has created a list of motifs with all information about them. These motifs are ranked by their occurences. The ouput is the same as the previous step but a classification is added

Plot the result

There is 3 ways to visualize the result.

Example

To see an example of output : Output Example