“…Each activity a i ∈ A is encoded as a vector (A i ) of length |A| + 3 such that the first |A| features are all set to zero, except the one occurring at the index of the current activity a i , which is set to one. Table 3 shows an example of prepared input sequences, where [1,2,3,4,5,6] are the resulting tokens or integers of the tokenized activities [N U LL, a 1 , a 2 , a 3 , a 4 , a 5 ], and the target column contains the encoded dummies (i.e. the converted categorical labels) [a 1 , a 2 , a 3 , a 4 , a 5 , EN D], where every activity is set to zero except the target activity, which is set to one.…”