A Methodological Approach to Model CBR-based Systems
JJournal of Computer and Communications, 2020, 08, 1-16
A Methodological Approach to Model CBR-basedSystems
Eliseu M. Oliveira , Rafael F. Reale and Joberto S. B. Martins Universidade Salvador - UNIFACS, Salvador, Brazil Instituto Federal da Bahia - IFBA, Valena, BrazilEmail: [email protected], [email protected], [email protected]
How to cite this paper:
Oliveira,M. E.; Reale, R. F.; Martins, J.S. B. (2020) A Methodological Ap-proach to Model CBR-based Sys-tems, Journal of Computer andCommunications, 08, 1-16.https://dx.doi.org/10.4236/jcc.2020.89001.2020.89001
Received: July 22, 2020Accepted: September 01, 2020Published: September 04, 2020
Copyright c (cid:13)
Abstract
Artificial intelligence (AI) has been used in various areas to supportsystem optimization and find solutions where the complexity makes itchallenging to use algorithmic and heuristics. Case-based Reasoning(CBR) is an AI technique intensively exploited in domains like man-agement, medicine, design, construction, retail and smart grid. CBRis a technique for problem-solving and captures new knowledge by us-ing past experiences. One of the main CBR deployment challenges isthe target system modeling process. This paper presents a straight-forward methodological approach to model CBR-based applications us-ing the concepts of abstract and concrete models. Splitting themodeling process with two models facilitates the allocation of exper-tise between the application domain and the CBR technology. Themethodological approach intends to facilitate the CBR modeling pro-cess and to foster CBR use in various areas outside computer science.
Keywords
Artificial Intelligence; Case-based Reasoning; CBR Modeling; Bandwidth Allo-cation Model.
1. Introduction
Artificial Intelligence (AI) and Machine Learning (ML) techniques are be-ing extensively used in an ever-increasing number of areas and systems.They provide benefits on adopting them, such as efficient optimizationmethods and the possibility to solve rather complex multi-objective andmulti-constrained problems that were difficult or eventually impossible to
DOI: 10.4236/jcc.2020.89001 Sep. 04, 2020 a r X i v : . [ c s . A I] S e p . M. Oliveira et al solve with current algorithmic or heuristics solutions [1].ML-assisted applications are a trend, and many researchers and devel-opers are rushing to apply ML and recover their inherent potential benefits[2] [3].However, using ML techniques to solve any problem do require someprevious background and expertise. For example, it is vital to choose theML technique that better suits the target application in terms of availablecomputational capability and expected target results. In sequence to anadequate ML technique choice, it is typically necessary to model the prob-lem under the premises of the chosen technique. The modeling processmay include, as an example, an MDP-based markovian process (MarkovDecision Process) like Q-Learning or SARSA formulation for Reinforce-ment Learning or the definition of a neural network structure for NeuralNetworks (NN) [4] [5].Case-based Reasoning (CBR) [6] is a technique for problem-solving andfor capturing new knowledge (learning) based on the stored knowledge ofpast experiences. CBR paradigm has a base of past experiences, called acase-base, and attempts to solve new problems by recovering similar solu-tions in this database and adapting them to new problems. CBR, to someextent, mimics the human behavior in activities like management and di-agnostics in which the previous knowledge and experience is the driver inlooking for the solutions for new near-equivalent problems [7] [8].CBR was proposed more than a decade ago as an AI technique [6], issimple to use, has minimal learning requirements, and does not typicallyrequire intensive computational resources [9]. More recently, CBR receivesthe attention from the artificial intelligence community and is gaining trackin domain areas like medicine, expert systems, retail, smart grid, construc-tion, manufacturing, design, agriculture and management [10] [11] [12] [13][14] [15].Like other AI techniques, CBR requires the target system to be mod-eled to allow a similarity search in its database. This process is not clearlydetailed or methodologically described in the literature. Our approach pro-vides a way to methodologically model the target system. In our approach,the CBR modeling for problem-solving requires a specialized abstract modelof the target system and, derived from it, a concrete CBR representationof the variables and parameters involved in the process.The objective of this work is, in summary, to propose a methodologi-cal approach to model the CBR process based on a mapping between theabstract and concrete representations of the CBR process variables andparameters. The proposed method aims to facilitate the CBR modelingprocess and contribute to promoting the widespread use of CBR. We alsoexpect the contribution can be relevant to CBR application areas wherethe computer science expertise of the professionals involved is less frequentor even unavailable.The remaining of this paper is organized as follows. Section 2 presentsan overview of the CBR fundamentals and section 3 discusses the relatedworks. Sections 4 and 5 present the CBR modeling methodology and sec-tion 7 follows with an example of how to use the approach for the cognitivemanagement of bandwidth in network links. In section 8, the final consid-erations are presented. DOI: 10.4236/jcc.2020.89001 Sep. 4, 2020 Journal of Computer and Communications . M. Oliveira et al
2. Case-based Reasoning Fundamental Aspects
In summary, in CBR, a new problem is solved through the knowledge andinformation available or acquired by a previous similar problem adapted tocreate a new solution.The essential components used to solve a problem with CBR are a case , the case-database of past cases or past experiences, and the similarityfunction .The case is the way we represent the experience we have about thetarget problem. All previous cases representing the acquired knowledgeare stored in a database, called case-database . A case is represented by apair problem and solution that are the fundamental aspects present in allCBR systems: • Problem : It contains all the information regarding the past event thatyou want to remember. That is, it describes all the essential datafor the representation of knowledge in the specific domain. Thesedata can include contextualization data, application objectives, de-scriptions of what happened, qualitative data, and quantitative data,among others. • Solution : It presents the necessary information to solve the problemrelated to the past event. This solution can be any information oraction that totally or partially solves the problem presented in thecase description. The representation of the solution must always takeinto account the application domain.The execution of the CBR process includes 4 phases using its essentialcomponents, namely (Figure 1) [16] [17]: • Recover : Performs the search for similar cases in the case-database.This comparison is made using the similarity function. The similarityfunction is responsible for comparing the base cases with the actualproblem and should return the most similar cases found. • Reuse : The phase in which the description of the current problemis composed with the solution of the case recovered in the previousphase. Then the solution found is applied to the environment inquestion. • Revision : In the review phase, the specialist in the technological do-main must assess whether the solution employed brings the expectedresults. The specialist has the opportunity to make fine adjustmentsto optimize, adjust, or adapt the recommended solution. • Retention : After making the necessary adaptations, the specialistmust confirm the new case as a valid case and consequently save it inthe case base for later reuse.From the methodological point of view, the utilization CBR to solve aparticular problem involves the following modeling and operational steps: • Case and Knowledge Representation : The case components and itsstructure embed the experience and the knowledge for a particularproblem. As such, case descriptions document real experiences, andthe CBR system may acquire new knowledge by retaining new cases.
Journal of Computer and Communications DOI: 10.4236/jcc.2020.89001 Sep. 4, 2020 . M. Oliveira et al
Figure 1: CBR 4R Cycle [16] • Similarity Measure : A similarity function calculates the similaritymeasure, and it defines to what extent a case in the case-database issimilar to the case being processed. • Adaptation : In CBR, the operational adaptation step takes a similarcase and adapts it to the current situation. In general, adaptationuses a defined mechanism, expert knowledge, or a mix of both. • Learning : It is an operational step that allows the CBR system tomemorize its successful and unsuccessful solutions, which means ef-fectively to acquire new knowledge.The foundation of the CBR that supports its capability to adapt, learn,and retain knowledge are the case representation and the choice of similaritymeasures. Accurately modeling these two elements of the CBR operationis essential.In this paper, we present a methodological approach that addresses therepresentation of cases and the choice of similarity measures for the problemof resource management with CBR.The proposed methodological approach is composed of an AbstractModel (AM) and a Concrete Model (CM). The abstract model is a high-level representation that structures the knowledge corresponding to thescope of actuation. The concrete model maps the abstract model’s repre-sentation to the set of parameters used in the CBR execution process toacquire knowledge for a set of particular cases.
3. Related Work
Watson presents a discussion about the methodological approach used byCBR in [18]. Watson argues that CBR can be better described as a method-ology for problem-solving and, as such, differs from other artificial intelli-gence techniques like neural networks, and genetic algorithms that use moreformal mathematical methods. Watson’s paper does not discuss or proposea methodological approach to support CBR deployments.
DOI: 10.4236/jcc.2020.89001 Sep. 4, 2020 Journal of Computer and Communications . M. Oliveira et al
The case representation formalism is discussed in Martines [19]. The au-thors discuss how experiences can be represented by using since simple fea-ture vectors to representational formalisms like object-oriented, predicate-based and semantic nets, among others.Althof discusses in [20] the need to develop CBR applications as a sys-tematic engineering activity. The paper addresses the software developmentlife-cycle targeting the development of software products that support CBRdevelopment.A discussion about CBR modeling is presented in Krite [21]. The paperdescribes how CBR can be used to compare, reuse, and adapt inductivemodels that represent complex systems. The paper does not address specificcases and knowledge representations.
4. Abstract Model
The proposed approach to model a CBR-based system for problem-solvinghas two steps: • Abstract Model (AM) definition; and • Concrete Model (CM) mapping.The definition of the AM requires expertise in the application domaininvolved. The mapping to the CM from the AM requires a minimum CBRexpertise.The abstract model is a high-level representation to structure the knowl-edge and, as such, to define the scope of action of the problem-solving sys-tem. It requires specialist knowledge of the application domain to whicha system using CBR is applied. This model aims to represent knowledgeabout the scope of actions for the CBR system.The abstract model is composed by a Technological Domain (TD)with general and specific objectives, attributes, measurements, actions, andpremises.
The technological domain defines, in general, the scope of the targetproblem-solving or, in other words, what the problem is and how to repre-sent it [16] [17].The representation of the TD in the proposed model is as follows: • System and problem description and objectives; • Static, contextual and dynamic attributes of the system; • Measurements; and • Actions.This set of components adequately describes the target system for gen-eral resource management problems like virtual machine management, dat-acenter management, and network management to cite some applicationexamples.
Journal of Computer and Communications DOI: 10.4236/jcc.2020.89001 Sep. 4, 2020 . M. Oliveira et al
The system and problem description is a formal or textual description ofCBR problem and system.The general objective is a high-level definition and delimitation of thetarget management problem. It is associated to the CBR resource manage-ment task.The specific objectives detail the general objective of the CBR manage-ment system’s multiple possible outcomes. In effect, this corresponds to thedefinition of a multi-objective problem, possibly under a multi-requirementscenario.The specific objectives specify a subset of the context, measurement, andaction attributes used in the CBR decision-making and learning processes.
The static attributes describe static characteristics of the target system andhave, in most cases, documentary value. The CBR system operation doesnot index the static attributes, so they do not interfere in the problem-solving process.The contextual attributes indicate the context of the problem and cor-respond to definitions and parameters that do not change frequently. Fromthe CBR system perspective, contextual attribute modification may implyin restarting the learning process with CBR.The dynamic attributes are the set of variables and parameters thatindicate the target system’s global state. In the problem-solving context,dynamic attributes impact the target management process and are mea-surable.
The measurements are, in general, the set of variables actual values acquiredby a monitoring system that are relevant to the management process. Themeasurements, together with the context attributes, provide a snapshot ofthe system current state.The measurements are instances of the system variables and aims tohave a snapshot of the systems and quantify the specific objectives. Ineffect, the management of accepted values for measurable variables quantify,in general, the specific objectives.
Tolerance represents the accepted value for the scope of this method-ology. So, in most cases, the specific objectives are represented by ranges,upper or lower limits for defined managed variables.The actions are the operation set that is used to react upon the identi-fication of a problem. The set of actions defined is executed on the systemand are related to the defined specific objectives.
In the CBR abstract model, the premises are the set of problems to whicha solution is known. Premises are optional and aim to facilitate knowledgeacquisition by the CBR system. It is essential to highlight that the CBRsystem is capable of learning from scratch without any given premise.
DOI: 10.4236/jcc.2020.89001 Sep. 4, 2020 Journal of Computer and Communications . M. Oliveira et al
The premises can be confirmed or not, through the acquisition of newknowledge. Wrong premises will negatively impact the learning processsince the premise will be discarded to allow the acquisition of new knowl-edge.
5. Concrete Model
The concrete model corresponds to the mapping of the abstract modeldefinitions in CBR cases, similarity function, parameters, variables, andweighting priorities of the deployed CBR system.The mapping from the abstract model to the concrete model is achievedin the following way: • TD components like attributes, measurements, and tolerances aremapped in the CBR cases description. • General and specific objectives are mapped in similarity function andevaluation function. • Actions are mapped in solutions for the CBR cases. • Premises are mapped in the first cases for the CBR system.The proposed sequence of steps to map from AM to CM model is asfollows:1. Mapping of attributes and measurements necessary to achieve thespecific objectives;2. CBR case description definition using the mapped set of attributesand measurements;3. Tolerance level definition for the CBR case components;4. Mapping of the actions for the solution of CBR cases;5. Similarity function definition;6. Evaluation function definition; and7. CBR 4R cycle operation process.The last step, the CBR 4R cycle operation process, is described in thispaper as a non-exhaustive sequence of how to do it procedures and hints that can be used in the CBR 4R cycle.
In CBR terminology a case is a problem situation composed by a set ofparameters describing the problem domain and the associated solution forthe problem (Figure 2) [22].A generic CBR case is: C j = ( p j , a j ) (1) Journal of Computer and Communications DOI: 10.4236/jcc.2020.89001 Sep. 4, 2020 . M. Oliveira et al
Figure 2: A CBR case where, p j = { SA j , CA j , M j , T j } (2) a j = { a j , a j , ..., a jx } (3) SA j = { SA j , SA j , ..., SA jn } (4) CA j = { CA j , CA j , ..., CA jk } (5) M j = { M j , M j , ..., M jz } (6) T j = { T j , T j , ..., T jy } (7)The case C j is composed by the set of parameters p j and the associ-ated set of actions a j . The set of parameters p j includes all the relevantstatic attributes ( SA j ), context attributes ( CA j ), measurements ( M j ) andtolerances ( T j ) for the problem situation being described.The case solution is composed by the set of relevant actions a j to achievethe defined objectives. These actions allow operations or sets of operationsthat are used to react to a particular problem in the system. Similarity is a crucial aspect of CBR. A similarity function is used to re-trieve similar cases from the case-database when a new case or unsolved
DOI: 10.4236/jcc.2020.89001 Sep. 4, 2020 Journal of Computer and Communications . M. Oliveira et al case arrives at the system. In summary, a new case or unsolved case is thedescription of a new problem to be solved.Examples of similarity functions include the identification of similaritybased on rules, correlation testing, K-nearest-neighbor (KNN) techniques,and the cosine similarity measure, among others [19] [11] [18] [16] [23].The similarity function, regardless of the option, requires the definitionof the evaluation indexes and their weights for the given real-world problem.The definition of the similarity measures is an actual challenging researchproblem and has a significant dependency on the target problem [24].From the methodological point of view, the similarity function ismapped from the specific objectives and is composed of a set of attributesand measurements with weights that define their priority in the definitionof the similarity.The similarity of one case to others can, from the methodological pointof view, be defined by averaging the distinct similarities of part of the case,for example, a case with three indexed attributes ( x i , x j , x l ) will be similarto another case ( y i , y j , y l ) if the attributes i , j and l of the cases x and y are similar to each other. This partial similarity, by attribute, is calledlocal similarity.The local similarity is calculated according to the type of attribute thatdefines the case, and a specific function can be used for each type of data.The functions used for calculate local similarity, as an example, are ladderfunction, linear function, equality function, maximum function, intersectionfunction, and contrast function [16].The global similarity determines how similar one case is to the otherusing the values of local similarities to which weights can also be assigned.Attributes and measurements have direct and indirect relation with thespecific objectives. Consequently, two issues arrive in terms of methodolog-ically mapping of the similarity function parameters: i) to consider indirectsimilarities; and ii) to consider similar cases in distinct contexts. Indirectrelations must be indexed, and a similar case in distinct contexts must bedifferentiated by choosing attribute weights adequately. The evaluation function is an optional facility that may be included in theCBR operation process. It is not conventionally inserted in the CBR cyclebut might be helpful in the CBR operation cycle. The EV is mapped fromthe objectives, attributes, and tolerances of the concrete model.The fundamental idea of using an evaluation function is to interpret thestate of the measurements, comparing them with the tolerances defined forthe CBR system. It provides a kind of on-the-fly evaluation of the CBRsystem behavior.This function can be used for two purposes: i) to generate periodicwarnings and diagnoses of symptoms and alerts detected in the CBR sys-tem; and ii) to check if a solution applied to the CBR system meets theobjectives (general and specific) and tolerances defined.
Journal of Computer and Communications DOI: 10.4236/jcc.2020.89001 Sep. 4, 2020 . M. Oliveira et al
6. CBR Operation Process - Hints and How to DoIt
After the mapping from the abstract model to the concrete model with thedefinition of objectives, attributes, measurements, actions, and cases, theCBR 4R cycle starts (Recover, Reuse, Revision and Retain).For each of these steps, some procedures shall be executed by the CBRsystem. These procedures depend on the actual problem-solving issue fo-cused on the CBR system, but there are commonalities. We explore someof these commonalities to provide a set of hints and possible how to doactions for the operation of CBR systems.
The Cycle 4R has as its starting point the recovery phase that retrieves asimilar case and evaluates it. This task can be triggered in two differentways: reactively or proactively.In reactive mode, an alarm requests analysis triggered by a current prob-lem, to obtain a solution or optimize the system. In proactive mode, theCBR system is activated to check the system’s situation and occasionallypropose improvements proactively or a solution.When no case returns from the case-database, two procedures are sug-gested: • Use the method assisted by the manager, where he provides a newsolution. • Use an automated method where the solution is automaticallymapped considering the defined objectives or an arbitrary solutionis attributed to the current problem. The arbitrary solution attribu-tion corresponds to a brute force learning method.
In the review phase, CBR assesses the efficiency of the proposed solution.For the review, it is necessary to wait a specific time until the actions takeeffect, and the attributes and measurements of the new state of the systemare updated.In this step, the evaluation function is used to check if the new solutionpresents improved performance. Without improvements, the new case isconsidered unsuccessful. The new case, positive or negative, is, by conven-tion, stored in the CBR database.After applying the solution, the current and previous attributes andmeasurements, cannot differ much from the previous state. With a con-siderable variation, it is not possible to identify whether the system hasimproved due to the solution adopted or simply because the state of the re-sources has changed. As adopted in other AI techniques, to use a discountis recommended in these cases.Another recommendation is the creation of a configurable equivalencethreshold for the cases. The aim is to avoid many very similar cases pop-ulating the case-database and contribute to its excessive growth, whichresults in a performance problem.
DOI: 10.4236/jcc.2020.89001 Sep. 4, 2020 Journal of Computer and Communications . M. Oliveira et al
False-positive cases can be negated in the next reuse. For false-negativecases, we suggest a period of validity, both for positive and negative cases,simulating the human forgetfulness of old facts that are rarely used.
7. Using the Abstract and Concrete Models for anCBR-based Application - Cognitive Managementof Bandwidth in Network Links
We exemplify now how abstract and concrete models can facilitate thedefinition of the objectives, attributes, measurements, and tolerances tomodel a CBR-based application.The target CBR-based application (BAMCBR) aims to manage band-width in links of an MPLS (MultiProtocol Label Switching Network). Thelink management is executed by a Bandwidth Allocation Model (BAM)that dynamically receives requests for link setup and grants or denies theserequests based on the link bandwidth availability in the network [25] [26].The cognitive management consists of CBR deciding when should the BAMmodel be changed among a set of options based on network parameters sta-tus. Oliveira in [27] has a detailed description of this cognitive managementapplication, and, in this paper, we focus on illustrating how AM and CMcan be used to model the CBR application.It follows the Abstract Model definition and Concrete Model mappings.
The first step in building the abstract (AM) model is the representation ofthe technological domain with the definition of its objectives, attributes,measurements, and tolerances.
The CBR system’s technological domain is the cognitive management of anMPLS/DS-TE type computer network with bandwidth allocation models(BAMCBR Tool) [27].
The general objective of the BAMCBR is to decide when should the BAMmodel be changed among a set of BAM model options available based onnetwork link-state performance parameters and input traffic.The specific objectives drive the BAM model reconfiguration decisionprocess and are the following:1. To maximize link throughput2. To minimize link preemption; and3. To minimize link devolution.It is essential to highlight that only with the expertise about theapplication domain (BAM model operation [26]) it is possible to knowthat throughput maximization, preemption, and devolution minimizationare management objectives achievable by the reconfiguration of the BAMmodel [28].
Journal of Computer and Communications DOI: 10.4236/jcc.2020.89001 Sep. 4, 2020 . M. Oliveira et al
The static attributes are the BAM model static configuration parameters.In the BAMCBR they are the total managed link bandwidth and the con-figured bandwidth allocate per class of traffic in the link (BC - BandwidthConstraint [26]).
The contextual attributes define the management context, and the modifi-cation of these attributes, although less frequent, does reflect the manager’sperception of what he wants from the CBR system. In BAMCBR, theadopted BAM and the tolerances for throughput, preemption, and devolu-tion are the main contextual parameters. Any change on these attributesimplies in restarting the CBR learning process.
The measurement attributes are the link variables that indicate the link’sperformance and state in a given moment. These are measurable dynamicvariables belonging to the target CBR system (network link). Examples ofmeasurement attributes used by the BAMCBR tool are link preemption,link devolution, packet loss, LSP (Label Switched Path) request blocking,and link utilization [27].
The tolerances represent the accepted values range for attributes in general.Their definition requires expertise in the domain area and they refines thespecific objectives. In the BAMCBR, as an example, the managementaccepts a link utilization of 10% with a 10% tolerance.
The actions correspond to the set of operations used to solve a problem fora certain case. In BAMCBR, the action is to reconfigure the current BAMmodel among the available options: MAM (Maximum Allocation Model),RDM (Russian Dolls Model) and ATCS (AllocTC-Sharing) models [28].
The premises are the set of problems or states to which a solution is known.For example, when the current BAM is MAM or RDM, and the utilizationis less or equal to 50 %, the BAM might be reconfigured to ATCS [28].
The concrete model (CM) maps the abstract model in the parameters andvariables used by the CBR system.
The case description includes the attributes, measurements and tolerancespreviously defined in the abstract model. In BAMCBR, it includes BAM
DOI: 10.4236/jcc.2020.89001 Sep. 4, 2020 Journal of Computer and Communications . M. Oliveira et al models, measurement variables like preemption, utilization, loss, blockingand devolution, and the action options for BAM model change (Figure 3).Figure 3: Caso BAMCBR
The actions set for BAMCBR is to dynamically switch between ATCS,RDM and MAM models [28].
The evaluation function verifies the actual state of the managed systemby comparing the actual case with previous cases using measurements andtolerances defined. For example, it must also be able to verify if a solutionapplied to the network is closer to the objective than the solution previouslyadopted. BAMCBR uses the WkNN function to evaluate the current stateof the network in relation to other previous states.Weights in the evaluation function reflect management expertise. Forexample, BAMCBR considers that devolution generates a negative impactmore significant than preemption, which, in turn, generates a more signifi-cant impact than blocking. As such, the weights for devolution, preemption,and blocking are 3, 2, and 1, respectively.
The functions for local and global similarity defined for the BAMCBR toolare indicated in Table 1.The cut-off threshold is used for recovering similar solutions within sim-ilarity equal or superior to the indicated limit. The equivalence thresh-old limit is used to avoid storing multiple nearly identical solutions at the
Journal of Computer and Communications DOI: 10.4236/jcc.2020.89001 Sep. 4, 2020 . M. Oliveira et al
Table 1: BAMCBR Similarity FunctionsAttribute/Measurement Similarity Function WeightBAM Model Equal Function 40Throughput Linear Function 30Blocking Linear Function 30Devolution Linear Function 20Preemption Linear Function 20Global Similarity Function: WkNNCut-off Threshold 96%Equivalence Threshold 98.5%database and, thus, populated it excessively. In other words, solutions withsimilarity equal or superior to the equivalence threshold limit are not storedin the CBR database.
8. Final Considerations
The methodological approach to model CBR-based applications uses thedefinition of an abstract model (AM) that is subsequently mapped in aconcrete model (CM).The abstract model represents the domain to which CBR is applied andits definition needs essentially the knowledge of an expert in the applicationdomain. The concrete model corresponds to the CBR parameters whosemapping from the abstract model requires CBR expertise. The splittingin two models facilitates the model process and, in addition, allows theallocation of domain and CBR specialists to the distinct phases of themodeling process.As such, this proposal’s inherent advantage is that it allows a taskdivision between specialists in the domain and specialist in CBR. This fa-cilitates the modeling process and has the potential to foster the utilizationof CBR in an even large number of areas where computer science expertiseis less frequent.
References [1] Raouf Boutaba, Mohammad A. Salahuddin, Noura Limam, Sara Ayoubi,Nashid Shahriar, Felipe Estrada-Solano, and Oscar M. Caicedo. A Compre-hensive Survey on Machine Learning for Networking: Evolution, Applicationsand Research Opportunities.
Journal of Internet Services and Applications ,9(1):16, June 2018.[2] Amina Adadi and Mohammed Berrada. Peeking Inside the Black-Box: ASurvey on Explainable Artificial Intelligence (XAI).
IEEE Access , 6:52138–52160, 2018.[3] Mohammad Saeid Mahdavinejad, Mohammadreza Rezvan, MohammadaminBarekatain, Peyman Adibi, Payam Barnaghi, and Amit P. Sheth. MachineLearning for Internet of Things Data Analysis: A Survey.
Digital Communi-cations and Networks , 4(3):161–175, August 2018.
DOI: 10.4236/jcc.2020.89001 Sep. 4, 2020 Journal of Computer and Communications . M. Oliveira et al [4] Richard S. Sutton and Andrew G. Barto.
Reinforcement Learning: An Intro-duction . Adaptive computation and machine learning. MIT Press, Cambridge,Mass., nachdr. edition, 1998.[5] German I. Parisi, Ronald Kemker, Jose L. Part, Christopher Kanan, and Ste-fan Wermter. Continual Lifelong Learning with Neural Networks: A Review.
Neural Networks , 113:54–71, May 2019.[6] Agnar Aamodt and Enric Plaza. Case-based reasoning: Foundational issues,methodological variations, and system approaches.
AI Commun. , 7(1):39–59,March 1994.[7] Petra Perner. Case-Based Reasoning and the Statistical Challenges II. InDr. Aleksandra Gruca, Tadeusz Czachrski, and Stanisaw Kozielski, editors,
Man-Machine Interactions 3 , Advances in Intelligent Systems and Comput-ing, pages 17–38, Cham, 2014. Springer International Publishing.[8] Ali Boukehila and Nora Taleb. Case-Based Approach to Detect Emergence. In
Proceedings of the 2019 3rd International Conference on Big Data Research ,pages 98–102, 2019.[9] Malik Jahan Khan, Hussain Hayat, and Irfan Awan. Hybrid Case-Base Main-tenance Approach for Modeling Large Scale Case-Based Reasoning Systems.
Human-centric Computing and Information Sciences , 9(1):9, March 2019.[10] Arkadiusz Weglarz and Pawe Gilewski. Application of CBR Systems in theProcess of Energy Retrofit of Single-Family Detached Houses.
MATEC Webof Conferences , 196:02033, 2018. Publisher: EDP Sciences.[11] Lukasz Osuszek and Stanisaw Stanek. Case Based Reasoning as an Element ofCase Processing in Adaptive Case Management Systems.
Annals of ComputerScience and Information Systems , 6:217–223, 2015.[12] Yikun Su, Shijing Yang, Kangning Liu, Kaicheng Hua, and Qi Yao. Devel-oping A Case-Based Reasoning Model for Safety Accident Pre-Control andDecision Making in the Construction Industry.
International Journal of En-vironmental Research and Public Health , 16(9), May 2019.[13] Nabanita Choudhury and Shahin Ara Begum. A Survey on Case-based Rea-soning in Medicine. In
Proceedings of the International Journal of AdvancedComputer Science and Applications , volume 7 of , pages 136–144, 2016.[14] Hugo Lopez-Fernandez, Florentino Fernandez Riverola, Miguel Reboiro-Jato,Daniel Glez-Pena, and Jose R. Mendez. Using CBR as Design Methodologyfor Developing Adaptable Decision Support Systems. Efficient Decision Sup-port Systems - Practice and Challenges From Current to Future , 2011.[15] Flvio G. Calhau and Joberto S B Martins. A Electric Network Reconfigura-tion Strategy with Case-Based Reasoning for the Smart Grid. In
Proceedingsof the VIII Brazilian Conference on Intelligent Systems (BRACIS) , pages 1–6,Salvador, Brazil, October 2019.[16] Eliseu M. Oliveira, Rafael F. Reale, and Joberto S. B. Martins. Evaluat-ing CBR Similarity Functions for BAM Switching in Networks with Dy-namic Traffic Profile. In
Proceedings of the V International Workshop onADVANCEs in ICT Infrastructure and Services , pages 1–7, Paris, January2017.[17] Aldo Von Wangenheim, Christiane Gresse von Wangenheim, and ThiagoRateke.
Raciocnio Baseado em Casos - 2 ed. Revisada e Atualizada . 072013.[18] I. Watson. Case-Based Reasoning Is a Methodology Not a Technology.
Knowledge-Based Systems , 12(5):303–308, October 1999.
Journal of Computer and Communications DOI: 10.4236/jcc.2020.89001 Sep. 4, 2020 . M. Oliveira et al [19] Zhaoyu Zhai, Jos-Fernn Martnez Ortega, Victoria Beltran, and Nstor Lu-cas Martnez. An Associated Representation Method for Defining AgriculturalCases in a Case-Based Reasoning System for Fast Case Retrieval.
Sensors ,19(23):5118, January 2019.[20] Ralph Bergmann and Klaus-Dieter Althoff. Methodology for Building CBRApplications.
Case-Based Reasoning Technology , pages 299–326, 1998.[21] Rosina Weber, Jason M. Proctor, Ilya Waldstein, and Andres Kriete. CBRfor Modeling Complex Systems. In
Case-Based Reasoning Research and De-velopment , pages 625–639, August 2005.[22] Jean-Baptiste Lamy, Boomadevi Sekar, Gilles Guezennec, Jacques Bouaud,and Brigitte Sroussi. Explainable Artificial Intelligence for Breast Cancer: AVisual Case-Based Reasoning Approach.
Artificial Intelligence in Medicine ,94:42–53, March 2019.[23] Kuo-Sui Lin. A Case-Based Reasoning System for Interior Design Usinga New Cosine Similarity Retrieval Algorithm.
Journal of Information andTelecommunication , 4(1):91–104, January 2020.[24] Nahyun Kwon, Kwonsik Song, Moonseo Park, Youjin Jang, Inseok Yoon, andYonghan Ahn. Preliminary Service Life Estimation Model for MEP Compo-nents Using Case-Based Reasoning and Genetic Algorithm.
Sustainability ,11(11):3074, January 2019.[25] Rafael Freitas Reale, Walter da Costa Pinto Neto, and Joberto S. B. Martins.AllocTC-Sharing: A New Bandwidth Allocation Model for DS-TE Networks.In
Proceedings of the VII Latin American Network Operations and Manage-ment Symposium , pages 1–4, Quito, Equador, October 2011.[26] Joberto Martins, Romildo Bezerra, Rafael Reale, and Gilvan Dures. Uma VisoTutorial dos Modelos de Alocao de Banda como Mecanismo de Provisiona-mento de Recursos em Redes IP/MPLS.
Revista de Sistemas e Computao ,5(2):144–155, December 2015.[27] Eliseu M. Oliveira, Rafael F. Reale, and Joberto S. B. Martins. CognitiveManagement of Bandwidth Allocation Models with Case-Based Reasoning -Evidences Towards Dynamic BAM Reconfiguration. In
Proceedings of theIEEE International Symposium on Computers and Communications (ISCC) ,pages 397–493, Brazil, June 2018.[28] Rafael Reale, Romildo Bezerra, and Joberto Martins. A Preliminary Evalua-tion of Bandwidth Allocation Model Dynamic Switching.
International Jour-nal of Computer Networks and Communications , 6(3):131–143, May 2014.
DOI: 10.4236/jcc.2020.89001 Sep. 4, 202016