Context-aware Adaption of Software Entities using Rules
Lauma Jokste and Jānis Grabis
Information Technology Institue, Riga Technical University, Kalku 1, Riga, Latvia
Keywords: Software Entity, Context-awareness, Rule-based Adaptation, Adaptation Requirements.
Abstract: Context-aware systems gain recognition in rapidly growing information systems market. Systems run time
adaption based on contextual information have been considered as a powerful mean towards better systems
performance which help to reach overall organizational goals and to improve key performance indicators.
This paper describes the concept where information systems can be divided into many software entities and
each of them can be context dependent. Context situation dependent software entity execution routines are
observed and these observations are used to formulate Context dependency rules either manually or by
machine learning. Rule based adaptation allows to monitor adaptation process in a transparent way and
allows to take into account human knowledge in adaptation process. The entity based adaption allows for a
uniform approach inducing context-dependency to different part of the software.
1 INTRODUCTION
Enterprise applications such ERP systems are used
to execute business processes. They provide a large
number of different functions and are characterized
by high degree of complexity. These applications are
typically packaged applications and often need to be
tailored for needs of specific users. However, their
complexity reduces modification flexibility, and
achieving the right fit between functionality provi-
ded and user needs is challenging. As the result, the
enterprise applications do not always yield expected
benefits (Calisir and Calisir, 2004) and suffer from
poor usability (Singh and Wesson, 2009).
Adaption of applications during their usage is
one of the solutions to improve performance of
enterprise applications (Macías-Escrivá et al. 2013).
Existing research towards adaptive enterprise
applications is limited to specific types of
adaptation, such as user centric (Shakshuhi et al.,
2015; Lavie and Meyer, 2010), data base
recommendations (Rubens et al., 2011), workflow
adaption (Uchibayahi et al., 2012), process-based
service provisioning (Pernici, 2007). In the case of
enterprise applications with a wide-scope such an
approach is infeasible because of the effort
associated with development of adaptive behavior.
However, enterprise applications are well-structured
and consist of standardized objects such as classes,
forms and reports. Each of them can be described by
several characteristics. These objects are referred as
to software entities (SEs) and adaptive behavior
could be standardized as a set of typical adaption
actions performed with the SEs. Adaption is usually
performed in response to changes in operating
circumstances and is aimed to satisfy specific goals
(Šūpulniece and Grabis, 2015). The operating
circumstances can be represented by contextual
information where context is defined as any
information affecting the SE (Dey, 2001). In this
paper, it is assumed that adaption is invoked in
response to changes in context.
The objective of this paper is to introduce the
concept of context-aware adaption of SEs
constituting enterprise application and to describe
the overall architecture of Context-aware software
entity adaptation (CASEA) system. Adaptive
behavior is defined using adaptation rules combining
SEs, context and adaptation actions. Adaptability is
delivered in a form of recommendations to avoid
obstruction of business process execution. Quality of
the adaption rules and recommendations made is
also evaluated. The main distinctive feature of the
proposed approach is that adaption is performed in a
uniform manner for all SEs constituting the
enterprise applications and adaption is externalized
without affecting development and maintenance of
key functionality.
The paper is structured as follows. Section 2
reviews related works and formulates requirements
166
Jokste, L. and Grabis, J.
Context-aware Adaption of Software Entities using Rules.
DOI: 10.5220/0006366401660171
In Proceedings of the 19th International Conference on Enter prise Information Systems (ICEIS 2017) - Volume 3, pages 166-171
ISBN: 978-989-758-249-3
Copyright © 2017 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved

towards the adaptive system. The overall
architecture of CASEA is described in Section 3.
Section 4 presents the structure of Adaptation rules.
Section 5 concludes.
2 FOUNDATIONS
The proposed research is motivated by a need to
improve efficiency of a suite of enterprise
applications used mainly by municipalities. These
enterprise applications provide a wide range of
services to a diverse group of users. The software
vendor has observed that there is relatively large
number of user support requests, some functionality
is used only to limited extent and many change
requests could be accommodated by existing
functionality. These issues could be addressed by
using adaptation. Additionally, there is a potential
for knowledge exchange between users, user groups
or municipalities because some municipalities have
significant in-house experience while others have
very limited information technology resources.
In order to identify ways for providing
adaptability, related research is reviewed and
requirements are formulated according to the review
and industrial experiences.
2.1 Related Work
Design of an adaptive systems deals with multiple
issues. These include the overall design of the
adaptive system, context processing and definition
and execution of adaptation policies.
The design of adaptive systems is guided by the
classical MAPE architecture (Andersson et al.,
2009). This architecture distinguishes the managed
subsystem (i.e., enterprise applications), managing
subsystem (i.e., adaptation module) and
environment. The managed and managing sub-
systems jointly make up a self-adaptive software
system. The Music development framework
(Hallsteinsen et al., 2012) uses the MAPE approach
to create variants of a software application
depending on the current context. Adaption is
performed according to the runtime adaption model.
Adaption modules are also a part of runtime models
based approach for developing dynamically adaptive
systems (Loukil et al. 2017). Both approaches
decouple the adaptation module from other parts of
the system and focus on developing new adaptive
applications. Alferez et al. (2014) elaborates a
framework for adaptive service composition.
Adaption is performed by the Model-based
reconfigurator according to context information
using the Variability model, which defines various
features of the services used in composition and
serves as a basis for formulating adaptation policies.
Verification of the adaptation results in an important
part of the framework.
In the area of dynamically evolving information
systems adaptive and context-aware systems gain
rapidly growing popularity. Two basic forms of
context data are raw or low level context data
obtained directly from sensors and rich or high level
data where some qualitative data processing has
been applied (Dey, 2000). Just a single context
factor is often taken into account, e.g., location (Pils
et al., 2006), personal context (Dey, 2000). In
enterprise applications perspective combination of
context data can give a significant contribution.
Yan et al. (2012) have proposed an autonomic
middleware framework which focuses on qualitative
context processing. Ashbrook and Starner (2003)
were one of the first researchers introducing the
context prediction technique where users’ next
location was predicted gaining information from
GPS. Lately, context prediction also plays an
important role in context-aware adaptive information
systems by applying prediction algorithms such as
Collaborative-based Context Prediction (Voigtmann
et al., 2011).
Salehie and Ladan (2012) develop adaptive
applications by using the Goal-Action-Attribute
model. The model specifies that attributes are used
to measure performance of adaptive applications,
goals define expectations towards adaptation and
actions represent possible changes in software. The
actions can be of various type though the papers
focuses on autonomic computing. Association rules
(Tan et al., 2006) is another technique used for
defining adaptation policies. Decisions and behavior
represented by the association rules are often context
dependent. Association rules mining is used to find
relations between contextual information and
problems occurred in information systems (Liu et
al., 2008). Rules in adaptive systems can be defined
manually or generated automatically by machine
learning methods (Kolski et al., 1997). Yang et al.,
(2013) emphasize the importance of contextual
factors that affects users’ preferences. Context-based
inference rules discovered from association rules
mining are introduced which allows to identify
frequent associations between situation features and
relevant content characteristics (Ke and Liu, 2011).
Adaptation rules are also used as a basis for
executing adaptive actions focusing on business
process adaptation (Carvalho et al., 2015).
Context-aware Adaption of Software Entities using Rules
167

The aforementioned papers focus on association
rules mining techniques and lack information about
usage of obtained rules in general and specifically in
the case of enterprise applications. Also adaptation
process is oriented to specific software instances,
such as business processes adaptation. The approach
described in this paper focuses on the end-to-end
enterprise applications adaptation process starting
from identification of context dependent SEs,
defining context dependencies by association rules
mining and continuing with run-time adaption which
is made based on observed context situations and
corresponding context dependency rules. The
feedback about usage of the rules is also collected to
improve credibility of the defined rules.
2.2 Requirements
The surveyed related work provides basis for
formulating requirements towards the CASEA
approach. Also the authors experience with
municipalities’ information systems and users’
requirements towards a better usability of the system
has been taken into account. The requirements are:
1. The MAPE loop is used to guide the adaptation
with clearly identifiable managed and managing
sub-systems as well as context monitoring
system.
2. The adaption module is decoupled from the core
parts of the enterprise applications to avoid
modification and disruption of the key business
functions. It should be scalable to enable
processing of large data amounts.
3. Adaptation process should not focus on specific
software instances and should be applicable for
different kind of SEs following a uniform design.
4. Behavior of SEs is altered by a set of well-
defined adaption actions to ensure predictable
adaption results and to reuse adaptation actions
across different SEs.
5. The adaptation actions are triggered by changes
in application execution context including
business and technical context and a context
module is used for pre-processing contextual
data coming from various sources.
6. The adaptation process should be able to learn
from itself and information about application
usage should be exchanged across the platform
to achieve knowledge sharing efficiencies.
7. Adaptation should be done in an unobtrusive
way for application users.
Given the size of enterprise applications, the
requirements specifically target development, reuse
and scalability issues. Effort required for developing
and maintaining adaptive components should be
outweighed by adaption gains. The adaptive
components should be developed in a uniform
manner to reduce complexity.
3 CASEA ARCHITECTURE
It is assumed that application consists of software
entities of different granularity and type. The SE is
defined as an executable software artifact which can
be either a software as a whole or a single workflow,
procedure, job, data input field or even a data value
of classifier. Software division in SEs can be
represented as an object tree following the principles
of software ontology (Kiefer et al., 2007). A SE
from a lower level can inherit context dependencies
from the upper level.
There is an open set of adaptation actions which
can be performed on every entity. Adapting a system
at the SE level allows to adjust it to specific context
situation and specific user needs more precisely. The
architecture described in this section is based on the
Recommendations metamodel (Jokste, 2015). The
metamodel implies that each SE has a context set
which includes all the necessary information about
context affecting the entity.
In order to create the architecture supporting the
adaptation process based on contextual information,
several requirements have been determined by
analyzing existing enterprise applications execution,
users’ activity data and log files from audit protocols
by applying different data mining and association
rules mining techniques. Requirements coincide
with those summarized in Section 2.2.
We came to conclusion that CASEA should be
built as an independent data base module where new
incoming context sources and elements, Context
dependency rules and Adaptation rules could be
added or removed without changing software code
(Fig.1). The rules should contain information about
SE, such as type, name, application etc. The module
would allow to choose an appropriate adaptive
action for each SE for the certain context situation.
CASEA consists of three core parts – input part,
execution of the module and output part. The main
input data are contextual information which can be
received from different internal or external data
sources with a certain regularity. Received context
information is then saved in context data base. If raw
context data is received, context element values
normalization process should be done in order to
convert context data into usable format. Replication
mechanism is defined for context database. Each
ICEIS 2017 - 19th International Conference on Enterprise Information Systems
168

time when new context information is received,
replication mechanism calls the rules compliance
checking procedure which checks if there exists
Context dependency rules which correspond to the
received context element values. If compliance is
found, then the appropriate Adaptation rule is
executed. Execution of Adaptation rules can be also
initiated by users’ action, e.g. if user enters the
system, replication mechanism checks the current
context information and verifies compliance with
Context dependency rules.
Figure 1: Context-aware SE adaptor module.
It is important to gather knowledge about the
quality of the defined or generated rules. That is
achieved by continuous feedback monitoring process
based on which rules weights can be recalculated.
The feedback monitoring process includes recording
users’ response to adjustments made.
4 ADAPTATION PROCESS
Software consists of a set of SE which can be
designated as SE, thereby the software (set of SE)
can be designated as SE = {se
1
, se
2
, se
3
, …, se
n
}.
Each SE can be dependent of one or many context
element one or many values. One context element
can affect one or several SE. All context elements
affecting the context-aware software are designated
as CE = {ce
1
, ce
2
, ce
3
, …, ce
n
}.
CASEA is rule-based and includes two different
kinds of rules. Context dependency rules define
context ranges which affect the SE and Adaptation
rules which initiate adjustments appropriate for the
certain run-time context situation. Context
dependency rules can be expressed as context-based
association rules where execution of SE related
adaptive action associates to a particular context
situation. Adaptation rules can be expressed as
event-condition-action (ECA) rules and designates
the adaptation action in a context situation.
Due to the fact that context element values
should be taken into account when defining Context
dependency rules, association rules should be
extended by including context element values, thus
simple Context dependency rule is defined as se
n
⇒
ce
n
(v), where se
n
∈ SE and ce
n
∈ CE, meaning that
SE se
n
associates with context element ce
n
, where
variable v indicates the context element value or
values interval of the context element measurable
property on which se
n
depends. The quality of the
rule is represented by weights RW = {rw
1
, rw
2
, …,
rw
n
}. The rule weight describes the impartiality of
one single or multidimensional Context dependency
rule. If one context situation is favorable to several
SE of the same kind, then SE with the highest
weight can be executed. The rule weights can be
recalculated in run-time based on several events,
such as the number of times rule compliance is
found, users’ feedback etc. Context dependencies
might consist of several context value combinations,
thereby multidimensional form of association rules
should be used.
Figure 2: Interrelationship between Context Dependency
and Adaptation Rules.
Adaptation rules define the adaptation task which
should be executed when context situation defined in
Context dependency rules occurs. For adaptation
rules we chose the form of ECA rules (also known
as reaction rules and post-condition rules) where
event is a run-time context situation, condition is
context element values (certain values or value
intervals) and action is an executable adjustment (SE
which can be executed in different ways). To ensure
quality of the adaptation process, the Adaptation rule
should include information about expected action of
the user when Adaptation rule is executed. If a user
has received a recommendation based on current
context situation to execute a certain process in
application, system expects the user to execute this
process. If the user follows the recommendation,
then it has been useful and thereby the Adaptation
rule is considered as apposite. Connection of
Context-dependency and Adaptation rules is shown
in Fig.2.
Context-aware Adaption of Software Entities using Rules
169

5 APPLICATION EXAMPLE
CASEA is intended for contextualizing large scale
enterprise applications. In this section we
demonstrate a shot example case from the Single
Municipalities System (SMS) developed by the
Latvian company ZZ Dats Ltd. SMS includes
different mutually integrated applications providing
IS support for implementation of municipalities’
functions and public electronical services for
citizens and companies.
Figure 3: Fragment of software division into SEs
Fig.3 demonstrates a fragment of SMS division
into SEs. Context dependency and Adaptation rules
are defined for the selected highlighted SEs. Context
elements affecting the execution of SE should be
identified. Both – personal and situational context
can be taken into account. For instance, lessee type
can affect the execution of several SE related to rent
objects. If number of lessee agreements >10 and
total rented area > 10000 m
2
, then lessee profile is
“active”. If Context-dependency rules are defined
manually then the initial rule weights are also
defined. Context dependency rules examples:
1) E-service: Real estate rent object list
⇒ lessee
profile (‘active’)
2) List column: area
⇒ lessee profile (‘active’)
3) Search field
⇒ time spent in object list(>180 sec)
4) Publish online procedure
⇒ unsuccessful searches
per hour (>10)
Adaptation rules examples:
1) IF lessee profile=’active’ THEN highlight rent
objects in list WHERE rent object area≥30000 m
2
2) IF lessee profile=‘active’ THEN automatically
order list by area column descending
3) IF time spent in object list >180 sec THEN
highlight search field
4) IF unsuccessful searches per hour>10 THEN
automatic e-mail/text notification to RENT user/-s.
When Adaptation rule is executed, users’
response is expected, for instance, if lessee profile is
‘active’ and rent object list is automatically ordered
by area descending, then the e-service user is
expected to apply for object with a large area or at
least to open the detailed information of object with
large area. If CASEA systems feedback monitoring
process identifies that the user executes the expected
action than the corresponding association rule
weight is increased. The execution of expected
action is fixed in audit protocols, thus CASEA
system can compare the users executed action with
predefined expected action.
6 CONCLUSIONS
In this paper we have introduced a conceptual
architecture of CASEA which allows to adjust
enterprise applications at the SE level based on
contextual information. The adaptation process is
carried out in a form of rules execution thus
allowing to account for human factors. This paper
demonstrates the work in progress and marks the
future tasks and challenges. Approach conceptually
described in this paper will be further evolved and
applied for enterprise applications that are used by
municipalities of Latvia. Context data will be mainly
extracted from users’ action logs, error protocols and
also calendar events. Future work includes the
following aspects: (1) To analyze the SEs context
dependencies based on municipalities systems users’
action logs; (2) to expand the method for defining
context dependency and adaptation rules; (3) to
develop the method for adaptation process validation
and evaluation of its effectiveness. (4) to develop the
CASEA module; (5) to validate and evaluate the
CASEA module and adaptation approach which is
conceptually described in this paper.
ACKNOWLEDGEMENTS
The research leading to these results has received
funding from the research project "Competence
Centre of Information and Communication
Technologies" of EU Structural funds, contract
No. 1.2.1.1/16/A/007 signed between IT
Competence Centre and Central Finance and
Contracting Agency, Research No. 1.6
“Development of Software Adaptation Algorithms
and Module Based on Context Information
Extracted from Users Action Logs”.
REFERENCES
Alferez, G.H., Pelechano, V., Mazo, R. and Salinesi C.,
ICEIS 2017 - 19th International Conference on Enterprise Information Systems
170

2014. Dynamic Adaptation of Service Compositions
with Variability Models. Journal of Systems and
Software, 91, 24-47.
Andersson, J., de Lemos, R., Malek, S., Weyns, D. 2009.
Reflecting on self-adaptive software systems. In: Proc.
of the ICSE Workshop on Software Engineering for
Adaptive and Self-Managing Systems, pp. 38–47.
Ashbrook D., Starner T., 2003. Using GP to Learn
Significant Locations and Predict Movement Across
Multiple Users. Personal and Ubiquitous Computing,
7(5), 275-286.
Calisir, F. and Calisir F., 2004. The relation of interface
usability characteristics, perceived usefulness, and
perceived ease of use to end-user satisfaction with
enterprise resource planning (ERP) systems.
Computers in Human Behavior, 20, 4, 505-515.
Carvalho, J.E.S., Santoro, F.M. and Revoredo, K., 2015. A
method to infer the need to update situations in
business process adaptation. Computers in Industry,
71, 128-143.
Dey, A.K., 2000. Providing architectural support for
building context-aware applications. PhD thesis,
Georgia Institute of Technology.
Dey, A.K., 2001. Understanding and Using Context. In
Personal Ubiquitous Computing, vol 5(1), pp. 4–7.
Hallsteinsen, S., Geihs, K., Paspallis, N., Eliassen, F.,
Horn, G., Lorenzo, J., Mamelli, A., Papadopoulos,
G.A., 2012. A development framework and
methodology for self-adapting applications in
ubiquitous computing environments, Journal of
Systems and Software, 85, 12, 2840-2859.
Jokste, L., 2015. Towards a Model of Context-aware
Recommender Systems. In Proceedings of the CAiSE
2015 Forum at the 27th International Conference
CAISE, vol 1367, pp 145-152.
Ke C.K and Liu D.R., 2011. Context-based knowledge
support for problem-solving by rule-inference and
case-based reasoning. International Journal of
Innovative Computing, Information and Control, 7.
Kiefer, C., Bernstein, A. and Tappolet, J., 2007. Mining
Software Repositories with iSPARQL and a Software
Evolution Ontology. In Mining Software Repositories,
ICSE Workshops MSR ’07.
Kolski, C., Le Strugeon, E., Tendjaoui, M. 1993.
Implementation of AI techniques for “intelligent”
interface development. Engineering Applications of
Artificial Intelligence, 6, 295-305
Lavie, T. And Meyer, J., 2010. Benefits and costs of
adaptive user interfaces, Int. J. of Human-Computer
Studies, 68, 508-524.
Liu, D.R., Ke, C.K. and Wu M.Y., 2008. Context-based
Knowledge Support for Problem-solving by Rule-
inference and Case-based Reasoning. In Proceedings
of the Seventh International Conference on Machine
Learning and Cybernetics, pp 32015- 3210.
Loukil, S., Kallel, S., Jmaiel, M., 2017. An approach
based on runtime models for developing dynamically
adaptive systems. Future Generation Computer
Systems, 68, 365-375.
Macias-Escriva F.D., Haber R., Toro, R. and Hernandez
V., 2013. Self-adaptive systems: A survey of current
approaches, research challenges and applications. In
Expert Systems with Applications, 40(18), 7267-7279.
Pernici, B., 2007. Adaptive Information Systems.
Conceptual Modeling in Information Systems
Engineering. Eds. J.Krogstie, A.L. Opdahl and
S.Brinkkemper, pp.295-304.
Pils, C., Roussaki, I. and Strimpakou, M., 2006. Location-
Based Context Retrieval and Filtering, vol. 3987 of the
series Lecture Notes in Computer Science, pp 256-
273.
Rubens, N., Kaplan, D. and Sugiyama, M., 2011.
Recommender Systems Handbook: Active Learning in
Recommender Systems (eds. P.B. Kantor, F. Ricci, L.
Rokach, B. Shapira). Springer, pp 735-767.
Salehie, M., Tahvildari L., 2012. Towards a goal-driven
approach to action selection in self-adaptive software.
Software Practice & Experience, 42, 2, 211-233.
Shakshuhi, E.M., Reid, M., Sheltami, T.R., 2015. An
Adaptive User Interface in Healthcare, In: Procedia
Computer Science, vol 56, pp 49-48.
Singh, A. and Wesson, J., 2009. Improving the Usability
of ERP Systems through the Application of Adaptive
User Interfaces. In Proceedings of the 11
th
International Conference on Enterprise Information
Systems (ICEIS), vol SAIC.
Šūpulniece, I. and Grabis, J., 2015. Conceptual Model of
User Adaptive Enterprise Application. CSIMQ, 3, 84-
96
Uchibayahi, T., Bernady, O.A. and Shiratori N., 2012.
Towards an Adaptive Workflow with Multi-Agents in
a Semantic Grid. In 12
th
International Conference on
Computational Science and Its Applications, pp 20-25.
Tan, P.N., Steinbach M. and Kumar V., 2006. Introduction
to data mining: Association Analysis: Basic Concepts
and Algorithms, pp 327-414.
Voigtmann, C., Lun L.S. and Klaus D., 2011. A
Collaborative Context Prediction Technique. In:
Vehicular Technology Conference, IEEE, pp 1-5.
Yan, H., Zheng, D. and Wang, J., 2012. Research of
Quality Based Autonomic Context Processing for
Pervasive Applications. In 7
th
International
Conference on Computer Science & Education, pp
1234 – 1238.
Yang W., Liao, Q., Zhang, C., 2013. An Association Rules
Mining Algorithm on Context-Factors and Users’
Preference. In International Conference on Intelligent
Human-Machine Systems and Cybernetics, 190-195.
Context-aware Adaption of Software Entities using Rules
171