A Goal-Based Approach for Learning in Business Processes - Intentional Perspectives on Information Systems Engineering - page 247

4.3 Online Application
Learning can be performed periodically in an off-line manner. At runtime, new pro-
cess instances are created and executed. The learning results should be applied to
these new instances. Each new process instance should be classified to an existing
context group so path selection decisions can be made according to the recom-
mended path for the context group according to the improved process model. Some
decisions (e.g., assigning a worker or a machine) are usually made based on different
criteria (e.g., machine availability). However, the context group may set preferences
among possible paths (e.g., prefer a certain machine out of several available for a
given context group).
If a process instance cannot be classified to an existing context group (e.g., a new
product for a new market or some unfamiliar external event), decisions would be
made based on human (managerial) judgment. In all cases, the data of the process
instance, namely, context information, specific path, and outcome, will be stored in
the experience base repository and serve for future learning cycles.
5 Related Work
This chapter combines three issues that have so far been addressed separately with
respect to business processes, namely, goal orientation, context awareness, and
actual process paths. We claim that this combination is important in order to achieve
learning and improvement over time.
The business process research area has mostly focused over the years on control-
flow issues, while goals as the driving force of business processes have not been
extensively used. The conceptual basis for the work presented here is the Generic
Process model (GPM) framework [ 21, 22] , which relates to goals as a fundamen-
tal part of process specification. Relying on Requirements Engineering approaches,
GPM distinguishes hard goals of a process from its soft goals [ 19] . Incorporating
goals into process specification enables assessing the ability of a process to achieve
its (hard) goal, which is termed the validity of its design [ 20] . Understanding and
explicitly specifying process goals has also been shown to be a key issue for process
flexibility [ 18] . A similar perspective of goal orientation has been suggested by [ 4] ,
but their approach deals with hard goals only.
Another approach that addresses goal oriented business process modeling is pre-
sented in [ 13] , proposing the map representation [ 16] as an intentional process
specification. Map representation has been assigned GPM-based semantics in [ 19] ,
which highlighted the synergy gained by combining these two.
Context awareness of business processes has recently gained the attention of the
scientific community. The main efforts have been devoted to identifying the relevant
context of a process, to its representation in a model, and to articulating how it can
affect the process execution at runtime. Context identification has mainly been done
in a qualitative manner (e.g., using an onion model [17] ), while the algorithmic