Design and Simulation of a Bio-inspired Hyper-heuristic Generic Model for Solving Combinatorial Optimisation Problems

Design and Simulation of a Bio-inspired Hyper-
heuristic Generic Model for Solving Combinatorial
Optimization Problems
Sangeetha Muthuraman
Research Scholar,
Department of Computer Science and Engineering,
Manonmaniam Sundaranar University
14, V.O.C. Street,
Kamaraj Nagar, Puducherry - 605011
00918903864543
sangsasi@yahoo.com
V. Prasanna Venkatesan
Professor and Head,
Department of Banking Technology,
School of Management,
Pondicherry University, Kalapet,
Puducherry - 605014
00919486199939
prasanna_v@yahoo.com
Abstract—The approaches used in literature for solving combinatorial optimization problems have applied specific methodology or a
specific combination of methodologies to solve it. However, less importance is attached to modeling the solution for the given problem
systematically. Modeling helps in analyzing the various parts of the solution clearly, thereby identifying which part of the methodology
or combination of methodologies applied is efficient or inefficient. In order to find how efficient the different parts of the applied
methodology is or methodologies are, it may be better to solve the given problem using the notion of hyper-heuristics. This can be done
by solving the different parts of the given problem with many different methodologies realized, implemented and benchmarked,
enabling to choose the best hybrid methodology. A theoretical model or representation of the problem’s solution may facilitate clear
proposal and realization of the different methodologies for the various parts of the solution. The literature reveals that there is a need
for a generic model which could be used to represent the solution for combinatorial optimization problems. Therefore, inspired by the
basic problem solving behavior exhibited by animals in their day to day life, a new bio-inspired hyper-heuristic generic model for
solving combinatorial optimization problems has been proposed. To demonstrate the application of this generic model proposed a
problem specific model is derived that solves the web services selection/composition problem. This specialized model has been realized
with a trip planning case study and the results are discussed.
Keywords-hyper-heuristics; model for solving combinatorial optimisation problems; generic model; bio-inspired model; hyper-heuristic
model
I. INTRODUCTION
Numerous approaches are available to solve combinatorial optimization problems. This includes applying exact, heuristic,
meta-heuristic or hybrid meta-heuristic [7], [11]-[18], algorithms or techniques. All these approaches use some specific algorithm
or combination of algorithms to solve the given problem. That is, the solution to the given problem is frozen. However, it may be
better to solve the given problem by appropriately modeling its solution and implementing the modeled solution using the notion
of hyper-heuristics. Using a hyper-heuristic notion enables the various parts of the model to be implemented using different
algorithms, each of which can then be benchmarked. Benchmarking helps in comparing the performances of the algorithms
implemented for every part of the model and identifying which algorithm is most efficient for this part of the model. This may
enable in choosing an optimal combination which solves the problem best for almost all its instances, from the different
combinations of efficient algorithms. To achieve the above said goal systematically it is essential to represent the solution to a
problem in the form of a model, which may enable the understanding of the solution as well as the various parts that form this
solution better. Going one step further and making everything general, it may be better to have a generic model, which could be
used to represent the solution for combinatorial optimization problems, in an abstract manner from which problem specific models
could be derived.
The literature reveals that there is a need for such a generic model, which can clearly represent the problem’s solution and
increase the understanding of the strategies used to solve the problem in these hybrid arrangements. Therefore, inspired by the
problem solving behavior exhibited by animals in their day to day life, by selecting appropriate actions which result in appropriate
behavior during appropriate situations, a new bio-inspired hyper-heuristic generic model for solving combinatorial optimization
problems has been proposed. As this proposed model is a generic model it only gives a general representation of the concept;
Different problem specific models could be derived by using the concept of this generic model as a base. In this work, a specific
model to solve the web services selection/composition problem has been derived from the generic model that is proposed and its
correctness is demonstrated using a case study.
The rest of this paper is organized as follows. Section 2 presents the background for the proposed work and Section 3 presents
the motivation behind this proposed work. Section 4 discusses the newly proposed bio-inspired hyper-heuristic generic model for
solving combinatorial optimization problems. Section 5 describes the application of the proposed generic model to get a specific
model for the web services selection/composition problem. Section 6 presents the implementation details of the specific model by
considering a trip planning web services selection/composition scenario as case study. Section 7 discusses the results of the case
study. Section 8 concludes the paper and suggests some future enhancements.
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 15, No. 8, August 2017
170 https://sites.google.com/site/ijcsis/
ISSN 1947-5500

(IJCSIS) International Journal of Computer Science and Information Security,
Vol. XX, No. XX, January 201X
II. BACKGROUND
The literature reveals that there are a variety of techniques used by researchers to solve combinatorial optimization problems
[7], [11]-[18]. Since these problems are NP-Hard, recently researchers have adopted hybrid meta-heuristic approaches to solve
such problems. Therefore, an extensive study has been made on using hybrid meta-heuristic approaches for solving combinatorial
optimization problems [7] by collecting research resources between the years1990 - 2015. These research resources using hybrid
meta-heuristic approaches for solving combinatorial optimization problems seen in literature could be categorized into three
different categories [7] based on the work presented in the research resources studied.
The first category is the proposal of a hybrid meta-heuristic technique or methodology that can be used to solve combinatorial
optimization problems. Only a few papers proposed belong to this category and in most of these papers the authors have proposed
cooperative schemes between meta-heuristics and exact, heuristic or meta-heuristic methods [32], and then some authors have
analyzed, every scheme proposed with one or more examples [30]. The idea behind is to solve the given problem in the best
possible manner, which has led to such cross-fertilization produced by combining different optimization techniques. Some authors
have categorized these hybridizations [21].
The advantage of such hybrid meta-heuristic technique or methodology proposal is that it demonstrates how the various
techniques can be combined, serving as an initial point for researchers planning to develop hybrid meta-heuristics. It also
illustrates such co-operations in a detailed manner which may help in further discussion and implementation in future papers.
The limitation of this category is that it is only technique or methodology proposal which was not implemented. Therefore,
there seems to be only a theoretical approach which constitutes mixing of different techniques or methodologies, whereas
practical implementation and the results are hypothetical. Also, no attempt is made to identify the contribution of different
techniques to the overall strategies performance.
The second category of resources had both a proposal and implementation of a hybrid meta-heuristic algorithm. Most of the
research resources studied in the survey belongs to this category. A few papers that belong to the second category include the
planning of medium-voltage power distribution systems [29], Hybrid meta-heuristics algorithms for task assignment in
heterogeneous computing systems [24], GELS-GA: Hybrid Metaheuristic Algorithm for Solving Multiple Travelling Salesman
Problem [16], A two-phase hybrid meta-heuristic for the vehicle routing problem with time windows [23]. To solve such
problems efficiently, the authors have proposed hybrid meta-heuristic algorithms which combines guided local search and large
neighborhood search [29] as in case of medium-voltage power distribution systems, a Hopfield neural network (HNN) is
combined with Genetic algorithm (GA) and simulated annealing (SA) as in case of the task assignment in heterogeneous
computing systems [24], a hybrid algorithm called GELS-GA is used in which of the global search capabilities of GA with that of
the local search of GELS algorithm has achieved optimality as in Multiple Travelling Salesman Problem [16]. In addition, some
problems had more than one objective functions as in case of the vehicle routing problem with time windows which has two
objective functions [23]. The first one is the minimization of the number of vehicles (primary criterion) is achieved by means of
an evolution strategy in the first phase and the second one is the total travel distance (secondary criterion) which is minimized
using a tabu search algorithm in the second phase.
The advantage of this algorithm based approach is that since combinatorial problems were solved by implementing a specific
hybrid meta-heuristic algorithm, its performance could be analyzed when compared to using any exact, heuristic or meta-heuristic
approach which is used to solve the same problem. Most of the time significant improvements in performance were noted [16],
[23], [24], [29]. Efficient behavior and higher flexibility were significant benefits obtained by cleverly combining the different
algorithms [31] from both the meta-heuristic as well as outside the meta-heuristic field.
The limitations of this category include that there seems to be no modeling of the solution seen. Some of the authors
recommend that in future the researchers should make an attempt to have a sound scientific experimental methodology consisting
of theoretical models for describing the hybrid meta-heuristic algorithm used, as done in natural sciences [21]. Such models would
bring more clarity in terms of solution representation especially when different combinations of hybrid meta-heuristics were used
to solve a given problem.
The third category presented an implementation of a framework for solving the problem. Only a few papers proposed belong
to this category. Examples of papers that belong to the third category include Hybrid Metaheuristics for Generalized Network
Design Problems [25], Hybrid Meta-heuristic Frameworks: A Functional Approach [15], which is a project in computational
biology which makes use of meta-heuristic methods to identify homology between proteins etc. The Generalized Network Design
Problems solved in this work includes the Generalized Minimum Spanning Tree Problem (GMSTP), the Generalized Traveling
Salesman Problem (GTSP), the Generalized Minimum Edge Biconnected Network Problem (GMEBCNP), and the Railway
Traveling Salesman Problem (RTSP). To solve the above said problems the authors have combined linear and integer
programming techniques like branch and bound, branch and cut, etc. with that of meta-heuristics which compute approximate
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

solutions but usually require significantly less runtime. In order to identify the homology between proteins [15] five major types
of meta-heuristic algorithms have been selected for hybridization. A framework and library of combinators to allow the
implementation and hybridization of these algorithms has been created.
The advantage of this approach is that it helps in achieving the benefit of hybridization to some extent. That is since a number
of libraries of algorithms or toolkits were created in the framework, it fastens the process of identifying or testing the different
combinations of hybrid meta-heuristics for the given problem with different combinations of parameter settings which is
otherwise a time consuming task.
The limitation of such framework based approaches include specificity in terms of the problem or problems that this
framework could be used to solve, limitations of the specific exact, heuristic and meta-heuristic algorithms implemented in the
framework, the language chosen for implementation etc. In addition, in framework based approaches there were no concrete
modeling of the solution for the given problem seen.
III. MOTIVATION
Although a significant growing interest has been seen in the field of hybrid meta-heuristics it is still in its basic stages and has
sufficient space for new developments and research [32]. In general, the literature comprised of either a theoretical methodology
proposal or some form of implementation (algorithm or framework). It is necessary to do a detailed examination of the problem
to solve, before the selection of an appropriate solution technique that could be used to solve it [22]. One clear and concise way of
examining the given problem and understanding its various parts is by depicting the problem in the form of a model. Therefore,
there is a need for having an experimental methodology consisting of theoretical models [21] of the solution which is then
implemented. This modeling gives better clarity of the proposed solution and may help in enhancing the understanding of the
different parts which constitute the proposed solution. The different parts of this solution could then be implemented using
different exact, heuristic, meta-heuristic or hybrid meta-heuristic algorithms, each of which could be benchmarked, thus enabling
the identification of an optimal algorithm for each part. Therefore, this enables in choosing the optimal hybrid combination that
solves the problem best for almost all its instances [7].
Thus, by using such a hyper-heuristic notion of solving each and every part of the problem with different implementations and
by combining these implementations to form several hybrid solutions, the full benefit of hybridization could be captured. The
biggest advantage of using a model is that it would correctly portray the different hybrid methodologies used to solve the same
problem from an architectural perspective. Therefore, inspired by the basic problem solving behavior used by animals in their day
to day problem solving, by selecting appropriate actions which result in appropriate behavior during appropriate situations, a new
bio-inspired hyper-heuristic generic model for solving combinatorial optimization problems has been proposed. This is described
in the next section.
IV. BIO-INSPIRED HYPER-HEURISTIC GENERIC MODEL FOR SOLVING COMBINATORIAL OPTIMISATION PROBLEMS
The core concept based on which this newly proposed bio-inspired hyper-heuristic generic model functions is animal behavior.
Behavioral ecology is the scientific study of animal behavior on the ecological as well as evolutionary basis particularly in natural
environments [4]. Any behavior is said to be a combination of one or more actions [10]. When an animal decides to exhibit a
particular type of behavior in a particular situation, then it has to decide the actions that it needs to exhibit in order to achieve the
behavior. So, behavior can be defined as a set of one or more actions made by individuals, organisms, systems or artificial entities
in conjunction with the other systems or organisms around as well as the physical environment [10]. The most important decision
that an animal has to make in a day to day basis is to choose the apt behavior for the apt situation. This in turn involves choosing
the appropriate actions that compose the behavior as well. Therefore, choosing an apt behavior and an apt set of actions that
composes it is basic and important for the survival of any animal.
The new bio-inspired hyper-heuristic generic model proposed uses the methodology followed by animals in enabling an
accurate decision making of choosing the apt behavior and an apt set of actions for the current situation. Accordingly, the model is
divided into three major components. They are the nervous system, motor system and the memory or information/knowledge
store. The nervous system is once again divided into the nature-nurture function and, allothetic and idiothetic controls [2].
 Idiothetic control: If the behavior or selection of actions of an animal for a particular situation depends purely
on the internally registered or stored information then this behavior is driven by idiothetic control.
 Allothetic control: If the behavior or selection of actions of an animal for a particular situation depends purely
on the external information then this behavior is driven by allothetic control.
Therefore, the behavior can either be driven by pure idiothetic control or allothetic control or a combination of both,
depending on the situation at hand and the problem to be solved. Both the allothetic and the idiothetic controls receive
information from the external environment. The allothetic control passes the external information as it is to the actions selector
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

and the behavior selector. The idiothetic control uses a nature–nurture function to select the appropriate internal information from
the brain or memory of the animal with respect to the current situation. The nature function is in charge of providing information
with regard to the genetic contribution in molding the behavior. The nurture function provides information with regard to the
learning and experience that an animal has gained from the environment [8], in modeling the behavior.
There are a lot of behavioral differences, seen from one animal to another, for the same situation, even if they belong to the
same species. This results due to the difference in experience. So, learning from experience is a major element that brings
behavioral flexibility, which is an important aspect which needs to be taken into account. Therefore, learning can be classified into
three types. They are environment based learning, learning of innate behavior and knowledge or experience based learning. The
animal acquires environment based learning from the external information that it gets from its environment. Social learning [1],
interactive teaching learning [3] and spatial learning and cognition [4], [5], could be categorized as part of environment based
learning. Innate behaviour is usually fixed and has strong genetic influence. Imprinting [4] could be categorized as part of innate
behavior learning.
When learning is based on the knowledge or experience that an animal has [6] and the understanding of what would happen by
selecting a particular behavior and set of actions, then this is termed as knowledge or experience based learning. Associative
learning [4] could be categorized as part of knowledge or experience based learning. Most animals learn all their ways of living,
communicating, searching, and foraging etc. as part of, environment based learning, learning of innate behavior and knowledge or
experience based learning. All these types of learning are stored in the animal’s memory or information/knowledge store which in
turn guides the nature-nurture function of the nervous system in choosing the appropriate information required or relevant to the
current scenario or situation. Thus, the external information from the allothetic control and the internal information from the
idiothetic control are passed to both the actions and behavior selector of the motor system. The actions and behavior selector does
the necessary processing and applies a suitable ranking algorithm on the internal or external information or a combination of both
to arrive at a set of actions and a corresponding behavior. Thus, the proposed model depicts how an animal makes fast and
accurate decisions with regard to appropriate actions and behavior selection. This logic followed by animals in apt behavior
selection could be used to solve several combinatorial optimization problems. The Fig. 1 below depicts a bio-inspired hyper-
heuristic generic model for solving combinatorial optimization problems.
External Ranked and
Information Selected
Behavior
External
Information
Ranked and Selected
Appropriate
Actions
Fig. 1. Bio-inspired Hyper-heuristic Generic Model for solving Combinatorial Optimization Problems
This is a generic model which has been adopted by animals. Depending on how complex the behavior of an animal is the
complexity of the algorithms used by the allothetic control, idiothetic control, nature-nurture function and combined
actions/behavior selector will also increase or decrease. Ranking algorithms are used in appropriate actions and behavior
Memory or information/knowledge store
(Environment based learning,
Learning of innate behavior and
Knowledge or Experience Based Learning)
Nervous system
Allothetic
Control
Idiothetic
Control
Motor system
Notify inappropriate
actions
Actions
selector
Nature-Nurture
Function
Behavior
Selector
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

selection, which ranks the various actions and behaviors and selects the best actions and behavior for the current situation. Thus,
the more accurate the ranking algorithms are, the more perfect will be the actions/behavior selection.
The next section details on the application of the proposed generic model to web services selection/composition problem.
V. WEB SERVICES SELECTION/COMPOSITION MODEL
An architecture that uses various mechanisms to solve the problem at hand with the help of a collection of services is basically
called as service-oriented architecture (SOA). The fundamental elements for developing applications in SOA are services. These
services converse with each other to solve the given problem. This conversation can entail either simple data passing between the
services or it could entail some kind of co-ordination of two or more services which thereby harmonize some activity. Therefore
the services are connected to one another by some means or the other. This architecture is actually based on two primary actors.
They are the service provider and the service requester or customer. The service provider, based on his assignment and available
resources, puts together pieces of executable software applications that provide some functionalities and non-functionalities; the
service requestor based on his service description, calls and executes such pieces of software that were put together as a whole in
order to solve the problem in hand. In real life, there are a lot of services available. Examples include train ticket booking, flight
reservation, banking, shopping, paying of household bills etc. It is very important for a service requestor to find and choose the
most apt and absolutely necessary services according to their requirements in case of an SOA based environment.
Today, there are a number of services available with the same functionality and this number is going to increase further in the
years to come. With this increase in number, the services with the same functionality are selected based on their non-functional
properties that suit the customer’s request [20]. Most of the time any task requested by a service customer is achieved by using a
combination of two or more services. For example train ticket booking may require the processing of ticket enquiry, ticket
payment and ticket reservation or booking services to co-ordinate with one another in order to achieve the task. Hence, the process
of web service composition is done by selecting the required services based on their functional and non-functional properties and
compositing them. Here one apt service is chosen from a set of available services for each activity and these selected services are
composed in such a way that the composition meets the consumer’s request. Therefore, one of the most important problems that
need to be dealt with is the web service selection/composition problem in case of service oriented architecture. To solve this
problem of web service selection/composition a specific web service selection/composition model [19] is derived from the
proposed bio-inspired hyper-heuristic generic model. The Fig. 2 below gives the web services selection/composition model
derived from the proposed generic model.
User service Ranked and
Preferences Selected
Compositions
User service
Preferences
Ranked and
Selected
Appropriate
Services
Fig. 2. Web Services Selection/Composition Model derived from the Proposed Generic Model
Knowledgebase and Database system that
stores the internal information details for
services, its selection and composition
Information/Knowledge system (for
web services selection and composition)
External
information/k
nowledge
passing
system
Internal
information
/knowledge
passing
system
Processing system (for
web services selection and
composition)
Notify inappropriate
services
Services
Selector
Information/Knowledge
selection function
Service
Composition
Selector
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

The architecture of this web services selection/composition model comprises of information/knowledge system,
knowledgebase and database system and the processing system. The information/knowledge system in turn consists of external
information/knowledge passing system, the internal information/knowledge passing system and information/knowledge selection
function. The customer’s preferences regarding the needed service are accepted from the customer by means of both external
information/knowledge passing system and the internal information/knowledge passing system.
1. Internal information/knowledge passing system: This system receives the customer or service user’s request. In
order to make appropriate decisions for web services selection and composition according to the current request this system
uses the Information/Knowledge selection function.
 Information/Knowledge selection function - The appropriate information/knowledge with regard to the
current user’s request is retrieved from the database and the knowledge base respectively by this
Information/Knowledge selection function.
2. External information/knowledge passing system: This system receives the customer or service user’s request.
This input is nothing but the customer or user preferences about the various functional and non functional properties of the
services requested, that is required for selection and composition processing. This system passes the current state of art
information and knowledge as it is to the services selector and composer.
Knowledgebase and Database system: This is used for storing web services related information, knowledge related to previous
web services selection and composition that could be reused etc.. The web services related internal information includes the
various functional and non functional properties of the services provided by the various service providers which are stored in the
service database or registry. The knowledge related to previous web services selection and composition is obtained from the
processing of previous similar or dissimilar customer requests. This previous processing results are thus stored in a
knowledgebase for reuse whenever necessary. This knowledge is used by the services selector and composer. Additionally,
customer feedback and rating for the services provided by different service customers’ is also valuable internal information that
could be stored and processed during service selection and composition.
Thus, the external information/knowledge passing system passes the request as it is and the internal information/knowledge
passing system helps in providing the various services related QOS information, foraged information or information based on
experience about the services and compositions stored in the database and knowledgebase. The appropriate information with
regard to the current request is obtained from the knowledgebase and database by the information/knowledge selection function.
Therefore the information/knowledge system helps in passing appropriate external and internal information required for further
processing to the processing system. The processing system consists of services selector and composition selector. These services
selector and composition selector selects the appropriate set of services and their corresponding composition by employing some
exact, heuristic, meta-heuristic or hybrid meta-heuristic algorithms. Any selection or composition decision made by the selector or
the composer depends on the information and knowledge passed by both the external as well as the internal
information/knowledge passing system. Under special circumstances either the internal information/knowledge passing system or
the external information/knowledge passing system may dominate, but in normal circumstances any decision made by the
processing system is a reflection of a combination of both.
VI. CASE STUDY AND IMPLEMENTATION
The specific web services selection/composition model described above is implemented by considering a trip planning
scenario as a case study for service composition [19]. This scenario consists of five different services with functionalities such as
hotel booking, airline reservation, train reservation, bus reservation and car rental services. The customer’s request any
combination of these services for planning a trip, by specifying the functionality and non-functional attributes as their user
preferences. The non-functional or QOS (Quality of service) attributes considered in this implementation includes price,
execution time, availability, reliability and the reputation of services. These functional and non-functional qualities are received as
input from the customers by both the internal and external information/knowledge passing system. The services selection and
composition is done by considering these functional and non-functional qualities.
The external information/knowledge passing system passes these inputs as it is to the services selector and composer whereas
the internal information/knowledge passing system uses the information/knowledge selection function to select the relevant
information/knowledge corresponding to the customer’s request from the information/knowledge store and passes this to the
processing system. In our implementation, the information/knowledge selection function uses simple database queries to select the
appropriate information/knowledge from the information/knowledge store. To do the service and composition selection according
to the customer’s request, the services selector has been implemented by using a reference score and trust based services selection
algorithm and the composition selector is implemented using a strategic tree based composition algorithm [19].
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

These algorithms are implemented on the Windows XP platform using Microsoft Visual Studio .NET development
environment and Microsoft Visual C# as the programming language. The service and QOS registries are implemented as a web
service which interacts with the Microsoft SQL Server 2005 database, which is used to store both the functional and non-
functional values, service customer requests and provider details. However, the limitations of this implementation include using
hard customer constraints and objective QOS values only. When using hard customer constraints, the services do not get selected
even if there is a minor deviation in the constraints requested by the customer. Therefore, it may be better to implement the same
algorithms by applying soft customer constraints. For simplicity, we have used only objective QOS values. It may be better to use
subjective QOS along with this objective QOS as this enables receiving soft customer’s constraints in a detailed manner, helping
customers to specify the level of requirement for all these QOS attributes and the level of relaxation for every QOS attribute when
the level of requirement cannot be achieved.
VII. RESULTS AND DISCUSSION
Although importance is attached to testing based web services ranking and reputation based web services ranking [28], so far
there seems to be no paper found in literature which has compared the selection algorithm’s performance based on the service
provider’s trust and customer satisfaction as an important factor. With the enormous increase in the number of web services it is
necessary to analyze the genuineness of the performance of services (in turn the genuineness of the service providers) which may
increase the satisfaction of customers. Hence, the parameters that were considered to evaluate the selection algorithm were the
number of successful service selection with and without considering trust, and customer satisfaction on the selected services when
selecting it with and without considering trust. Therefore, the selection algorithm was executed by both considering and without
considering the service providers trust factor [19]. That is the service providers were ranked based on the genuineness of the QOS
values provided by them for their services. The QOS values provided by the service providers were compared with the monitored
QOS values. The providers were ranked based on the amount of deviation. Lesser deviation indicates more genuineness. The Fig.
3 below depicts the number of successful service selections with and without considering providers trust for 30, 50 and 100
service requests.
30 service requests 50 service requests 100 service requests
28
45
94
20
39
82
Number of Successful Service Selections With and
Without Considering Service Provider's Trust
without trust with trust
Fig. 3. Successful selection of services with and without considering service provider’s trust
For 30, 50 and 100 service requests 28, 45 and 94 services are selected without considering trust and 20, 39 and 82 services
are selected with considering trust. Therefore, from the above figure it is clear that number of services selected is more when not
considering service provider’s trust as an important factor in selecting services and that number of services selected is less when
considering service provider’s trust as an important factor, in selecting services. But, the number of selected services that
customers were satisfied are 18 out of 28, 27 out of 45 and 56 out of 94 when not considering trust and 17 out of 20, 35 out of 39
and 74 out of 82 when considering trust. Therefore, although the number of services selected are said to be more the customer
satisfaction was as less as 59% in considering selection without trust. But, the customer satisfaction was 85% or more when trust
was considered, as almost the quality values provided by the service providers have matched the monitored quality values. Thus,
the Fig. 4 below depicts the percentage of customer satisfaction with and without considering providers trust for 30, 50 and 100
service requests.
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

30 service
requests
50 service
requests
100 service
requests
customer satisfaction
without trust 64.28% 60% 59.57%
customer satisfaction
with trust 85% 89.74% 90.20%
0.00%
10.00%
20.00%
30.00%
40.00%
50.00%
60.00%
70.00%
80.00%
90.00%
100.00%
CustomerSatisfaction
Percentage of Customer Satisfaction With and
Without considering Service Provider's Trust
Fig. 4. Percentage of Customer satisfaction with and without considering service provider’s trust
The composition algorithm was evaluated in terms of the number of enumerative compositions generated for a single request
while selecting the best composition. In general, the composition algorithm proposed performed better when compared to other
algorithms seen in literature [26, 27]. When the proposed composition algorithm was compared with a traditional algorithm [26],
in terms of the number of enumerative compositions generated for a single composition request of five abstract services as
considered in the case study. The proposed algorithm performed well as the number of enumerations it generated was only a
maximum of 243 compositions in the worst case [19], whereas the other algorithm generated a minimum of 3125 compositions or
even more in the worst case. In the best case only one single composition was generated and on an average analysis the number of
compositions generated was less than 100 most of the time. The Fig. 5 below depicts the comparison of the composition algorithm
based on number of enumerations for a single composition request in the worst case.
3125
243
Traditional composition algorithm Proposed composition algorithm
Comparison of the proposed composition algorithm with a
traditional algorithm in terms of number of composition
enumerations for a single composition request in the worst case
Fig. 5. Comparison of number of enumerations made for a single composition request by the proposed composition
algorithm with other traditional algorithms
VIII. CONCLUSION AND FUTURE ENHANCEMENTS
Any classic approach taken in modeling begins with a proposal of a theoretical model, with many new predictions based on
observation [9]. Therefore, in this paper a new bio-inspired hyper-heuristic generic model for solving combinatorial optimization
problems has been proposed based on the observed problem solving behavior of animals. As this model is a generic model it
could be realized into a specific model, according to the problem that needs to be solved. In this work the generic model has been
applied to obtain a specific model for the web services selection/composition problem. This specific model has been implemented
for a trip planning web services selection/composition case study and the results have been analyzed. Further, in future different
algorithms could be applied for the various elements or parts of the web services selection/composition model, thus identifying
the most optimal methodology for every part of the model, from which the most optimal hybrid methodology which solves the
problem efficiently for almost all its instances could be identified. In addition, the proposed bio-inspired hyper-heuristic generic
model could be applied in future to solve various other complex combinatorial optimization problems as well. The biggest
advantage of using a model is that it would correctly portray the different hybrid methodologies used to solve the same problem
by using the notion of hyper-heuristics from an architectural perspective, when the same problem is solved using different hybrid
methodologies.
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

REFERENCES
[1] Lawrence Williams, Martha Glasgow, “Squirrel Monkey Behavior in Research”, J. ILAR, vol.41, no.1, pp. 26-36, 2000.
http://ilarjournal.oxfordjournals.org/content/41/1/26.full
[2] J.H.Visser, “Host –Plant Finding by Insects: Orientation, Sensory input and Search Patterns”, J. Insect Physiol, vol.34, no.3, pp. 259-268, 1988.
[3] Wikipedia, the free encyclopaedia, “Ant - Behaviour and Ecology”, http://en.wikipedia.org/wiki/Ant#Behaviour_and_ecology, (accessed 1-9-2015).
[4] Erin Barley, Fathleen FitzPatrick, “Animal Behavior”, www.dublin.k12.ca.us/cms/lib5/.../192/Chapter%2051%20Animal%20Behavior.ppt
(accessed 1-9-2015).
[5] Susanne Åkesson, Rüdiger Wehner, “Visual navigation in desert ants Cataglyphis fortis: are snapshots coupled to a celestial system of reference?” The Journal
of Experimental Biology, vol. 205, pp. 1971–1978, 2002.
[6] eHow Contributor, “Knowledge Based Learning”, http://www.ehow.com/about_5403738_knowledge-based-learning.html, (accessed 1-9-2015).
[7] Sangeetha Muthuraman, Prasanna Venkatesan, “A Comprehensive Study on Hybrid Meta-Heuristic Approaches Used for Solving Combinatorial Optimization
Problems”, in Proceedings of the IEEE Xplore 2016 World Congress on Computing and Communication Technologies (WCCCT 2016), 7th
to 9th
December 2016,
Tamil Nadu, India. pp. 185-200, DOI 10.1109/WCCCT.2016.53.
[8] Andre Ferran, Anthony F.G. Dixon, “Foraging behaviour of ladybird larvae ( Coleoptera: Coccinellidae)”, European Journal of Entomol., vol. 90, pp. 383-402,
1993.
[9] Christian Blum, Andrea Roli, “Metaheuristics in Combinatorial Optimization: Overview and Conceptual Comparison”, ACM Computing Surveys (CSUR),
vol.35, no.3, pp. 268-308, 2003. http://kursinfo.himolde.no/forskningsgrupper/optimering/phdkurs/Metaheuristics%20in%20Combinatorial%20Optimization.pdf
[10] G¨unther R. Raidl, Jakob Puchinger2, “Combining (Integer) Linear Programming Techniques and Metaheuristics for Combinatorial Optimization”, Springer
Chapter : Hybrid Metaheuristics Volume 114 of the series Studies in Computational Intelligence 2008, pp. 31-62.
[11] Jorg Homberger, Hermann Gehring, “A two-phase hybrid metaheuristic for the vehicle routing problem with time windows”, European Journal of
Operational Research, Elsevier, vol. 162, pp. 220–238, 2005.
[12] Sancho Salcedo-Sanz,Yong Xub, XinYao, “Hybrid meta-heuristics algorithms for task assignment in heterogeneous computing systems”, Journal of
Computers & Operations Research, Elsevier, vol. 33, pp. 820–835, 2006.
[13] Bin Hu, “Hybrid Metaheuristics for Generalized Network Design Problems”, Ph.d thesis, Vienna University of technology, 2008,
https://www.ac.tuwien.ac.at/files/pub/hu_08.pdf (accessed 1-9-2015).
[14] Pieter Vansteenwegen, Wouter Souffriau, Kenneth S¨orensen, “Solving the mobile mapping van problem: A hybrid metaheuristic for capacitated arcrouting
with soft time windows”, Journal of Computers & Operations Research, Elsevier, vol.37, no.11, pp.1870-1876, 2010.
[15] Richard James Senington, “Hybrid Meta-heuristic Frameworks : A Functional Approach”, Ph.d thesis, The University of Leeds School of Computing,
February 2013, etheses.whiterose.ac.uk/4847/1/Richard%20Senington's%20thesis.pdf
[16] Ali A. R. Hosseinabadi, Maryam Kardgar, Mohammad Shojafar, Shahaboddin Shamshirband, Ajith Abraham, “GELS-GA: Hybrid Metaheuristic Algorithm
for Solving Multiple Travelling Salesman Problem”, in Proceedings of the 14th
Int. Conf. Intelligent Systems Design and Applications (ISDA), pp. 76-81, 2014.
[17] Thibaut Vidal, Maria Battarra, Anand Subramanian, Gunes Erdogan, “Hybrid Metaheuristics for the Clustered Vehicle Routing Problem”, Journal of
Computers and Operations Research, vol.58 no.C, pp. 87-99, 2015. http://arxiv.org/pdf/1404.6696.pdf
[18] Sherif A Masoud, Scott J Mason, “A bi-criteria hybrid metaheuristic for analysing an integrated automotive supply chain”, Journal of the Operational
Research Society, advance online publication, 30 September 2015, vol.67, no.3, pp. 516-526, 2016.
[19] Sangeetha Muthuraman, Prasanna Venkatesan, “Design of QOS based Web Service Selection/Composition Hyper-Heuristic Model”, in Proceedings of the
International Conference on Informatics and analytics (ICIA'16), 25th
– 26th
August 2016, Pondicherry Engineering College, Pondicherry, India, 2016, ACM,
ISBN: 978-1-4503-4756-3.
[20] Sangeetha Muthuraman, Prasanna Venkatesan, “Qualitative and Quantitative Review of QOS based Web Services Selection and Composition Techniques”,
in Proceedings of the International Conference on Informatics and analytics (ICIA'16), 25th
– 26th
August 2016, Pondicherry Engineering College, Pondicherry,
India, (2016), ACM, ISBN: 978-1-4503-4756-3.
[21] Christian Blum, Jakob Puchinger, Gunther R. Raidl and Andrea Roli, “Hybrid metaheuristics in combinatorial optimization: A survey”, Journal of Applied
Soft Computing, vol. 11, pp. 4135–4151, 2011.
[22] Van Valdhuizen D.A., Lamont G.B., “Multiobjective Evolutionary Algorithms: Analyzing the State-of-the-Art”, Evolutionary Computation, vol. 8, no. 2, pp.
125-147, 2000.
[23] Jorg Homberger, Hermann Gehring, “A two-phase hybrid metaheuristic for the vehicle routing problem with time windows”, Elsevier, European Journal of
Operational Research, vol.162, pp. 220–238, 2005.
[24] Sancho Salcedo-Sanz,Yong Xub, XinYao, “Hybrid meta-heuristics algorithms for task assignment in heterogeneous computing systems, Elsevier, Journal of
Computers & Operations Research, vol. 33, pp. 820–835, 2006.
[25] Bin Hu, “Hybrid Metaheuristics for Generalized Network Design Problems”, Ph.d thesis, Vienna University of technology, 2008,
https://www.ac.tuwien.ac.at/files/pub/hu_08.pdf
[26] Yuan-sheng Luo, Yong Qi, Di Hou, Lin-feng Shen, Ying Chen and Xiao Zhong, “A novel heuristic algorithm for QOS-aware end-to-end service
composition”, Journal of Computer Communications, vol.34, no.9, pp. 1137-1144, June 2011.
[27] Ping Wang, Kuo-Ming Chao and Chi-Chun Lo, “On optimal decision for QoS-aware composite service selection” Journal of Expert Systems with
Applications, vol. 37, no.1, pp.440–449, Jan 2010. DOI= https://ir.nctu.edu.tw/bitstream/11536/6100/1/000271571000050.pdf
[28] Vuong Xuan Tran, Hidekazu Tsuji and Ryosuke Masuda, “A new QOS ontology and its QOS-based ranking algorithm for Web services” Journal of Simulation
Modelling Practice and Theory, vol. 17, no. 8, pp. 1378-1398, September 2009.
[29] Xiaohu Tao, Hans-Jürgen Haubrich, “A HYBRID METAHEURISTIC METHOD FOR THE PLANNING OF MEDIUM-VOLTAGE POWER
DISTRIBUTION SYSTEMS”, 15th PSCC, Liege, pp. 22-26, August 2005.
[30] L. Jourdan *, M. Basseur, E.-G. Talbi, “Hybridizing exact methods and metaheuristics: A taxonomy”, European Journal of Operational Research, vol. 199,
pp. 620–629, 2009.
[31] Christian Blum, Andrea Roli and Michael Sampels, “Hybrid Metaheuristics: An Emerging Approach to Optimization” Studies in Computational
Intelligence, Springer 2008 edition, ISBN-10: 354078294X, ISBN-13: 978-3540782940.
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

[32] Christian Blum, “Hybrid Metaheuristics in Combinatorial Optimization: A Tutorial”, A.-H. Dediu, C. Mart´ın-Vide, and B. Truthe (Eds.): TPNC 2012, LNCS
7505, pp. 1–10, Springer-Verlag Berlin Heidelberg 2012.
Vol. 15, No. 8, Augus 2017
ISSN 1947-5500

Design and Simulation of a Bio-inspired Hyper-heuristic Generic Model for Solving Combinatorial Optimisation Problems

Recommended

More Related Content

What's hot (19)

Similar to Design and Simulation of a Bio-inspired Hyper-heuristic Generic Model for Solving Combinatorial Optimisation Problems (20)

Recently uploaded (20)

Design and Simulation of a Bio-inspired Hyper-heuristic Generic Model for Solving Combinatorial Optimisation Problems