3.2.6. Step six:
The index of the maximum value of each column in (\ref{707891}) is captured. The result is a vector \(v\) of size 3, named \(j_{max}\); simultaneously, the index of the minimum value of each column in (\ref{707891}) is captured. The result is a vector \(v\) of size 3, which is denoted as \(j_{min}\).
3.2.7. Step seven:
The two vectors in (\ref{985634}) are stacked; the result is a matrix of size \(2\times3\), named \(j_{img}\), and flattened to a vector of size \(1\times6\). As a result, each observation in __ds is transformed to a size of \(1\times6\). Finally, __ds is split into \(k\) subsets; each subset represents the individual traits of the class, and \(k\) corresponds to the number of classes in _ds. Fig. (\ref{908079}) illustrates the previous two steps using the showcase example in (\ref{993524}).
The __ds in (\ref{238699}), the \(v_{norm}\) in (\ref{238699}), the \(r_{inner}\) in (\ref{103702}), the avg vector in (\ref{840207}), the err vector in (\ref{790239}), and the individual traits in (\ref{662612}) are saved as the predictive model; Fig. (\ref{842432}) illustrates those elements.