Page 300 -

282 11 Analyzing “Lasagna Processes”

Fig. 11.4 Use cases for process mining combine goals (expressed in KPIs) and improvement
actions, e.g., process mining can be used to shorten the flow time by providing insights that lead to
a process redesign

• Identification of exceptional cases that generate too much additional work. By
learning the profile of such cases, they can be handled separately to reduce the
overall flow time.
• Visualization of the 10 most complicated or time consuming cases to identify
potential risks.

These use cases illustrate the potential of process mining. It is easy to imagine the
application of these use cases to the WMO process described earlier. For instance,
results such as shown in Fig. 11.3 can be used to discover bottlenecks and to gener-
ate ideas for flow time reduction. The results of conformance analysis as depicted in
Fig. 11.2(b) can be used to identify compliance problems, e.g., for the 32 cases hav-
ing missing or remaining tokens one could analyze the social network of the people
involved.

11.3 Approach

In Chap. 9, we described ten process mining related activities using the frame-
work shown in Fig. 11.5. These ten activities are grouped into three categories:
cartography (activities discover, enhance, and diagnose), auditing (activities detect,
check, compare, and promote), and navigation (activities explore, predict, and rec-
ommend). Although the framework helps to understand the relations between the
various process mining activities, it does not guide the user in conducting a process
mining project. Therefore, this section introduces the L life-cycle model for mining
∗
Lasagna processes.
Several reference models describing the life-cycle of a typical data mining/BI
project have been proposed by academics and consortia of vendors and users.
For example, the CRISP-DM (CRoss-Industry Standard Process for Data Mining)
methodology identifies a life-cycle consisting of six phases: (a) business understand-
ing, (b) data understanding, (c) data preparation, (d) modeling, (e) evaluation, and
(f) deployment [19]. CRISP-DM was developed in the late nineties by a consortium
driven by SPSS. Around the same period SAS proposed the SEMMA methodol-
ogy consisting of five phases: (a) sample, (b) explore, (c) modify, (d) model, and
(e) assess. Both methodologies are very high-level and provide little support. More-
over, existing methodologies are not tailored toward process mining projects. There-
fore, we propose the L life-cycle model shown in Fig. 11.6. This five-stage model
∗