Synthetic data for assessing and comparing local post-hoc explanation of detected process shift

Synthetic data for assessing and comparing local post-hoc explanation of detected process shift

DOI

10.5281/zenodo.15000635

Synthetic dataset contains data used in experiment described in article submitted to Computers in Industry journal entitled 

Assessing and Comparing Local Post-hoc Explanation for Shift Detection in Process Monitoring.

The citation will be updated immediately after the article will be accepted. 

Particular data.mat files are stored in a subfolder structure, which clearly assigns the particular file to

on of the tested cases. 

For example, data for experiments with normally distributed data, known number of shifted variables and 5 variables are stored in path \normal\known_number\5_vars\rho0.1.

The meaning of particular folders is explained here:

normal - all variables are normally distributed

not-normal - copula based multivariate distribution based on normal and gamma marginal distributions and defined correlation

known_number  - known number of shifted variables (methods used this information, which is not available in real world) 

unknown_number  - unknown number of shifted variables, realistic case

2_vars - data with 2 variables (n=2)

...

10_vars - data with 10 variables (n=2)

rho0.1 - correlation among all variables is 0.1

...

rho0.9 - correlation among all variables is 0.9

Each data.mat file contains the following variables:

  LIME_res      nval x n results of LIME explanation

  MYT_res       nval x n    results of MYT explanation

  NN_res        nval x n results of ANN explanation

  X                  p x 11000         Unshifted data              

  S                  n x n sigma matrix (covariance matrix) for the unshifted data              

  mu              1xn            mean parameter for the unshifted data             

  n                 1x1            number of variables (dimensionality)             

  trn_set         n x ntrn x 2    train set for ANN explainer, 

                      trn_set(:,:,1) are values of variables from shifted process

                      trn_set(:,:,2) labels denoting which variables are shifted 

                      trn_set(i,j,2) is 1 if ith variable of jth sample trn_set(:,j,1) is shifted

  val_set         n x 95 x 2          validation set used for testing and generating LIME_res, MYT_res and NN_res