Fig. 2.

Unfolding procedure to reduce the three dimension matrix to a two dimensional matrix. The initial dataset has batches (N) on the y-axis, variables (K) on the x-axis and time (J) on the z-axis in a three dimensional space. The aim of the unfolding is the same for both approaches, but the resulting two dimensional dataset is different. The final data set which will be used for the RDA (a) has a shape of NxJ rows with K columns. Each raw contains data points xijk from a single batch observation. The terminal data set which will be used for the FBA (b) has a shape of N rows with features (F) columns. A feature is a certain data information within a time series data. The number of features depends on the initial and final one point measurements as well as strongly on the amount of extraction done by the data scientist, which are evaluated by detecting deviations within a time series overlay plot