=Paper=
{{Paper
|id=Vol-2196/BPM_2018_paper_24
|storemode=property
|title=A Tool for the Uniqueification of DMN Decision Tables
|pdfUrl=https://ceur-ws.org/Vol-2196/BPM_2018_paper_24.pdf
|volume=Vol-2196
|authors=Kimon Batoulis,Mathias Weske
|dblpUrl=https://dblp.org/rec/conf/bpm/BatoulisW18
}}
==A Tool for the Uniqueification of DMN Decision Tables==
https://ceur-ws.org/Vol-2196/BPM_2018_paper_24.pdf
A Tool for the Uniqueification of DMN Decision
Tables
Kimon Batoulis and Mathias Weske
Hasso Plattner Institute, University of Potsdam, Potsdam, Germany
{Kimon.Batoulis,Mathias.Weske}@hpi.de
Abstract. The Decision Model and Notation (DMN) prescribes decision
tables as the standardized way of expressing decision logic. The rules of
the decision table may be overlapping, meaning that given some input,
multiple rules of the table match and some conflict resolution strategy has
to be applied to determine the final output. Therefore, the input-output
behavior of tables with overlapping rules is not immediately clear, which
makes them hard to understand and unsuitable for analysis tasks. In
this paper, we present a tool that transforms DMN decision tables with
overlapping rules into equivalent ones that only contain exclusive rules, a
process we call uniqueification.
Keywords: DMN, Decision Table Analysis, Uniqueification, Camunda
1 Introduction
The Decision Model and Notation (DMN) [7] is an OMG standard for the
design of decisions. In DMN, a decision is modeled on two levels. The decision
requirements level represents the dependencies between decisions and what input
data is necessary. The decision logic level, in turn, defines the logic of each
individual decision. DMN specifies decision tables as the default option for
that and prescribes a standardized design. An important attribute of a DMN
decision table is its hit policy. This attribute becomes important when tables
with overlapping rules are designed.
Consider the table in Fig. 1. This table assigns a credit rating score to a person
R Income (k) Assets (k) Credit Rating
Number ≥ 0 Number ≥ 0 A, B, C, D, E
1 ≤ 30 ≤ 30 A
2 [10..60] [10..25] B
3 [20..95] [40..90] C
4 ≥ 80 - D
5 ≥ 40 ≥ 85 E
Fig. 1. Example DMN decision table containing overlapping rules with a rule order hit
policy
based on their income and assets. However, for certain inputs more than one rule
matches. For example, for the input (15, 10) both rules 1 and 2 are eligible. This
is a conflict that requires further guidance, because it is not clear which rule’s
F. Casati et al. (Eds.): Proceedings of the Dissertation Award and Demonstration, Industrial
Track at BPM 2018, CEUR-WS.org, 2018. Copyright c 2018 for this paper by its authors.
Copying permitted for private and academic purposes. This volume is published and copyrighted
by its editors.
K. Batoulis, M. Weske
output should be chosen in this case. Therefore, a table with overlapping rules
has to indicate a hit policy that should be applied in such cases. This policy is
represented by a single letter in the upper left corner of the table. For instance,
the table in Fig. 1 specifies a rule order policy, represented by the letter R. With
this policy the outputs of all matching rules are collected in a list which is sorted
according to the order of the matching rules in the table. Therefore, the table
would return [A, B] as the decision result given the input (15, 10).
Apparently, the set of possible outputs of a decision table can be different
from the outputs of the individual rules, and depending on how many overlapping
rules there are, this set may not be obvious. For example, the set of possible
outputs of the example table is:
A, B, C, D, E, [A, B], [C, D], [C, E], [D, E], [C, D, E].
This set is not easily identifiable by looking at the table. The same holds for the
other way round. Which inputs are responsible for which outputs? Phrased more
generally, what is the input-output behavior of the table?
Knowing the input-output behavior of a table eases understanding its logic.
Also, it facilitates conducting certain analysis tasks that are based on the input-
output behavior, such as checking the sound integration of a decision table with
a process model [1, 2, 5]. Therefore, we developed and implemented an algorithm
that translates a decision table with any number of overlapping rules and any hit
policy into an equivalent one (i.e., one having the same input-output behavior)
containing only exclusive rules—a process we call uniqueification, which was
already described in [4]. In such a table the input-output behavior is clearly
visible from the individual rules.
2 Tool
The tool to uniqueify DMN decision tables is implemented in dmn-js, a DMN
decision table editor developed by Camunda1 . Our implementation builds upon
functionality provided in a tool to verify DMN decision tables [6] and also a tool
to check the soundness of decision-aware business processes [3]. In fact, it is an
extension of [3], with added functionality for table uniqueification.
Our tool is available for download together with exemplary models, the pseu-
docode of the algorithm, and a screencast at https://bpt.hpi.uni-potsdam.
de/Public/TableUniqueification.
In the following, we will apply the tool to the example from Section 1. Fig. 2
shows the interface of the tool after the table was opened. Since this is an
extension of an existing tool, there are several buttons at the top. The rightmost
one triggers the algorithm for uniqueifying the table. This will translate the table
into an equivalent one (in terms of input-output behavior) that now consists
only of exclusive rules. The view of the tool displaying this translated table is
shown in Fig. 3. The table has 15 rules, all of them being unique (i.e., exclusive).
1
https://camunda.org
A Tool for the Uniqueification of DMN Decision Tables
Fig. 2. View of the tool after opening a DMN decision table
Fig. 3. View of the tool after pressing the uniqueify table button for the table in Fig. 2
Therefore, the table’s hit policy is now changed to unique, denoted by the letter
U in the upper left corner.
3 Maturity
In this section, we discuss two aspects regarding our tool. First, we report the
results of our performance analysis of the uniqueification algorithm. Second, we
discuss the tool’s maturity.
For the performance analysis, we generated 1000 synthetic decision tables
with up to 50 rows and 30 columns. The results showed that there are two factors
that the runtime of the algorithm depends on: On the one hand, it is influenced
by the number of overlapping rules of the table. The more overlapping rules
there are, the more exclusive rules there can potentially be in the translated
K. Batoulis, M. Weske
table. For example, the maximum execution time of 151 seconds was reached for
a 50 × 6-table with 1138 overlapping rules. On the other hand, the runtime is
highly affected by the number of rows of the table because the more rows there
are, the more potential overlaps need to be checked. The average execution time
for the 1000 decision tables with up to 50 rows and 30 columns was 18 seconds.
Our tool is able to uniqueify any DMN decision table that is based on the
S-FEEL language. This language allows standard data types such as strings,
numbers and booleans to be used for the input values of the table. Moreover,
disjunctions of input values are permissible as well as multiple output variables.
The full FEEL language, however, that also supports, for example, formulating
output values as functions of input values (such as output = (input1+input2)×2)
is not supported.
References
1. Batoulis, K., Haarmann, S., Weske, M.: Various notions of soundness for decision-
aware business processes. In: Mayr, H.C., Guizzardi, G., Ma, H., Pastor, O. (eds.)
Conceptual Modeling. pp. 403–418. Springer International Publishing, Cham (2017)
2. Batoulis, K., Weske, M.: Soundness of decision-aware business processes. In: Carmona,
J., Engels, G., Kumar, A. (eds.) Business Process Management Forum. pp. 106–124.
Springer International Publishing, Cham (2017)
3. Batoulis, K., Weske, M.: A tool for checking soundness of decision-aware busi-
ness processes. In: Proceedings of the BPM Demo Track and BPM Disserta-
tion Award co-located with 15th International Conference on Business Process
Modeling (BPM 2017), Barcelona, Spain, September 13, 2017. (2017), http:
//ceur-ws.org/Vol-1920/BPM_2017_paper_184.pdf
4. Batoulis, K., Weske, M.: Disambiguation of dmn decision tables. In: Abramowicz, W.,
Paschke, A. (eds.) Business Information Systems. pp. 236–249. Springer International
Publishing, Cham (2018)
5. Hasić, F., Smedt, J.D., Vanthienen, J.: Augmenting processes with decision intelli-
gence: Principles for integrated modelling. Decision Support Systems (2017)
6. Laurson, Ü., Maggi, F.M.: A tool for the analysis of DMN decision tables. In: Proceed-
ings of the BPM Demo Track 2016 Co-located with the 14th International Conference
on Business Process Management (BPM 2016), Rio de Janeiro, Brazil, September 21,
2016. pp. 56–60 (2016), http://ceur-ws.org/Vol-1789/bpm-demo-2016-paper11.
pdf
7. OMG: Decision Model and Notation, Version 1.1 (May 2016)