Topic Case-Based Reasoning (CBR) and its Application in Bank and Medical Cases.
Dec 26, 2015
Topic
Case-Based Reasoning (CBR) and its Application in Bank and Medical Cases.
Brief description of CBR
CBR lifecycle and methodology
CBR merits
Review of CBR in Bank and medical
Open problems and research directions
Conclusions and references
Presentation outline
Case-Based Reasoning is a recognized and well established method for building medical expert systems and other expert knowledge based systems [3, 10].
CBR by a way of definition involves an approach to develop knowledge based systems that are able to retrieve and reuse solutions
Knowledge expert engineers design new expert system through the use of Case-Base, which implies re-using past similar problem solution cases as it’s applied to a new case.
A case-based reasoner solves new problems by adapting solutions that were used to solve old problems
Description of CBR
A methodology to model human reasoning and thinking
A methodology for building intelligent computer systems
– store previous experience (cases) in memory
– to solve new problems:
retrieve similar experience about similar situations from the memory
reuse the experience in the context of the new situation:
complete or partial reuse, or adapt according to differences
store new experience in memory (learning)
In a nutshell ,Case-Based Reasoning is ... :
Generally, A CBR cycle can be described in four processes; which are Retrieve, Reuse, Revise and
Retain [1, 2, 3] otherwise coined as 4R.
CBR Lifecycle
The first phase is to retrieve the most similar case or cases. Include ;
.By following direct index pointers from problem features
.By searching an index structure
.By searching in a model of general domain knowledge.
The second phase is to reuse the information and knowledge in that case to solve the problem.
The third phase is to revise the proposed solution. While,
The last phase is to retain the parts of the experience that are likely to be useful for future
problem solving.
– Select cases that can be adapted easily to the current problem
– Select cases that have (nearly) the same solution than the current problem
Efficient case retrieval is essential for large case bases
Reterival:Similarity matching in CBR Purpose of similarity:
– on the case representation
– size of the case base
• Organization of the case base:
– Linear lists, only for small case bases
– Index structures for large case bases
• Kd-trees: index structure for large case bases (Wess)
• Retrieval nets: index structure for textual CBR (Lenz)
• Discrimination nets: used with representations in logic
Different Retrieval Approaches Depends;
How to Adapt the Solution
- Different Options
No modification of the solution: just copy
Manual/interactive solution adaptation by the user
•Automatic solution adaptation
– Transformational Analogy: transformation of the solution
Reuse:
– Verification of the solution by computer simulation – Verification / evaluation of the solution in the real world
Criteria for revision
– Correctness of the solution
– Quality of the solution
– Other, e.g., user preferences
Revise: Verify and Correct Solution
What can be learned:
– New experience (new case)
– Improved similarity assessment, importance of features
– Organization/indexing of the case base to improve efficiency
– Knowledge for solution adaptation
– Forgetting cases, e.g., for efficiency or because out-of-date
• Methods
– Storing cases in the case base
– Deleting cases from the case base
– Explanation-based learning
– Induction, e.g. of decision trees
– Neural net style learning
Retain: Learning from Problem Solving
The CBR methodology is about the algorithm of interpreting and assimilating a new case from a previous case base.
The general CBR methodologies:
Assign Indexes: where the features of the new case are assigned as indexes characterizing the event.
Retrieve: where the indexes are used to retrieve a similar past case from the case memory
Modify: where the old solution is modified to conform to the new situation, resulting in a proposed
solution
Test: where the proposed solution is tried out. It either succeeds or fails.
CBR Methodology
Assign and Store: If the solution succeeds, then assign indexes and stores a
working solution. The successful plan is then incorporated into the case memory.
Explain, Repair and Test: If the solution fails, then explain the failure, repair the
working solution, and test again. The explanation process identifies the source of
the problem. The predictive features of the problem are incorporated into the
indexing rules knowledge structure to anticipate this problem in the future. The
failed plan is repaired to fix the problem, and the revised solution is then tested.
• Reduces the knowledge acquisition effort
• Requires less maintenance effort
• Improve problem solving performance through reuse
• Makes use of existing data, e.g. in databases
• Improve over time and adapt to changes in the environment
• High user acceptance
Advantages of CBR over other Techniques
CBR In Bank Case: Main issues: CBR approach for predicting bank lending decisions.
Economic agents/analyst find it hard in making decisions during bank lending cases, to interpret
credit result, and forecast the future behavior
the authors of [8] proposed a CBR approach that will help to mitigate the difficulty level in the
bank lending process.
In their study, they created a CBR system that would forecast the future behavior of
economic agents that offer credit.
The CBR proposed approach upon implementation saved the time and increase the speedy
process of making lending decisions
LITERATURE REVIEW…. CBR Application in Bank and Medical Cases
their CBR goal was to find solution to different problems such as:
Decision Making: foresee the decision of economic agents relating to the criterion
for loan approval and the conditions inherent to credit granting.
Decision Factors: These foresee relevant factors that determine changes in the behavior of
economic agents when granting credit
Economic sentiment: which details to what extent the expectation regarding economic
activity in general condition decisions.
GOAL;
The problem is specified as a set of objective factors relating to a given
situation in which the credit granting decision is to be made,
The hypothesis is a single condition representing the economic sentiment of the
decision-maker,
The solution is the decision regarding credit granting. It specifies a set of
actions relative to credit granting.
The generic case structure for their system took the following
approach; Description of main methodologies;
1.Check Case: ---------
Their CBR system was design to start by checking the case base relating to enterprises or households
as it pertains to a current case problem.
----The system will then search for the most similar cases to the problem currently described to the
system.
----The cases are selected when their similarity with the current problem exceeds the minimum level
of similarity specified to the system.
Two level of similarity; They experimentally took values of 0.7 as the lower limit for the first
similarity level and 0.5 for the second similarity.
2. Case Identification: The next process after the cases are selected is to identify the cases with the
highest degree of similarity. If one case is found the problem is solved by adapting the solution of the
selected case. If various cases were found, the economic analyst will have to decide the best solution.
Description of main methodologies:
when the best case or cases have been found, the condition attached to each case or cases were adapted
using the method of predicate instantiation.
the procedure for predicate instantiation uses synonyms in the form of synonym (P1, P2). This clause
means that predicate P1 and P2 are synonymous. The authors cited an example, where by the clause
synonym (market, behaviour, and demand for loans) means that the market behaviour is considered
synonymous with the demand for loans. ------Upon substitution of predicates,
the system again compares the problem with adapted cases, proposing the solution of the best case.
Hence, the solved problem becomes a new case, which is automatically added to the secondary case.
Predicate Instantiation
Another vital method in the proposed CBR approach is to compute the similarity of the two
cases, in other to pick/retrieve the best match.
calculating the similarity of two cases results from weighing the similarity of the condition
with the similarity of the hypothesis (the economic sentiment). The derived result from the
similarity calculation would be ;
the calculation of the similarity of the condition (S1)
calculation, if necessary of the similarity of the economic sentiment (S2)
calculation of general similarity (S): S = S1 * S2
Similarity computation
One of their main aims was to assess the role played by the economic sentiment in the decision making process of those granting
credit.
Upon the successful test and research on their system, the authors demonstrated through examples on how the system can be
used, taking on an economic case as the case example, the reply for the CBR approach are carried out through Prolog language;
An illustration on how the system can be used, with economic case is as follows;
-market behaviour is improving
-the cost of fund is increasing,
-competition between the bank is decreasing,
-There is a pessimistic economic sentiment as regards the economy in general.
Basic procedure carried out (1st criterion)
Solve _ Problem (enterprise, [market_behaviour (bank,increase),
cost_of_funds(bank,increase),pressure_from_competition(bank, decrease)],[economic_sentiment(bank,pessimistic)])
Result:
No solution found (1st criterion)
The procedure for the second, less restrictive criterion was carried out, as no solution was found,
Result of their experiment and tests:
2. Second procedure carried out (2nd criterion)
Solve2_ Problem (enterprise,[market_behaviour (bank, increase), cost_of_funds (bank, increase),
pressure_from_competition (bank, decrease)], [economic_sentiment (bank, pessimistic)]).
Result:
Problem
[market_behaviour (bank, increase),cost_of_funds (bank, increase),pressure_from_competition (bank, decrease)]
Economic sentiment
[economic_sentiment (bank, pessimistic)]
Best similar case
[debt_restructuring (bank, increase),
demand_for_loans (bank, increase),
cost_of_funds (bank, increase)],
[economic_sentiment (bank, pessimistic)],
[approval_of_loans (bank, tightened somewhat),
Spread (bank, tightened somewhat)]
Similarity: 0.75(2nd criterion)
Adaptation
[pressure_from_competition (bank, decrease),demand_for_loans (bank, increase),
cost_of_funds (bank, increase)],[economic_sentiment (bank, pessimistic)], [approval_of_loans (bank, tightened somewhat),spread (bank
The explanation of the above prolog illustration is as follows:
when the 1st (more restrictive) procedure was carried out, no solution was found;
when the 2nd (less restrictive) procedure was carried out using the 2nd similarity criterion, it was possible
to reach a result (forecast) through adaptation of the decision specified in the previously existing case that
has been selected.
The adaptation procedure uses the method of substitution by reinstatement of the predicate
Market behaviour by its synonym demand for loan. This allows for a similarity to be reached within the
criterion [0.7, 1], the interval defined for a solution to be considered. The similarity found (0.75)
corresponds to the difference of one predicate (pressure from competition and debt restructuring).
From the obtained result through the illustrated tests, the authors concluded that the system can forecast
with considerable estimate of 90%, of the decision of the economic agents. They also were of the opinion
that the experiments with the hypotheses also show that the economic sentiment play a decisive role in
both considered segment regarding the criterion for loan approvals and the other conditions underlying
credit granting.
Case 1: CANCER Diagnosis:
Salem, the author of [9] summarized the application of CBR application in cancer diagnosis which was developed by the medical
Informatics Research group at Ain Shams University.
The main aim of the reviewed CBR based experts system prototype for diagnosis of cancer diseases in [9] was to serve as doctor’s
diagnostic assistant. Also, the system was deployed as a tool to aid those suffering from intractable pains
The process of the CBR system design was based on the general knowledge engineering task in a CBR development. The system
approach consists of three main modules which are all interacted with the main environment of cancer diseases.
-- user interface ( Querying, Inference and Diagnostic case)
-- case base reasoning module( case retrieval, similarity and matches)
-- Computational module (rule based inference).
The patient cases are retrieved in dialogue with similarity matches using the nearest neighbour matching technique. As well, the frame
technique was used for patient indexing, storage and retrieval. The CBR initial process is done through examination of rules in the rule-
based inference. The rules would encode the patient’s information on symptoms and pathological examinations.
CBR in Medical Diagnosis
Main issues concerned: diabetes has a high prevalence risk, sometimes physicians taking care of
diabetic patients have no specialized formation in diabetes, consequently, the management of these
patients may be less accurate
Goal: proposed CBR application that would support an individualized assessment of the patients, so as to
improve both the management and the treatment of diabetes
Case II;Individual Prognosis Of Diabetes Long -Term Risk, By Armengol et al
DIRAS APPLICATION: presented an application know as DIRAS (Diabetes Individualized Risk Assessment System)
which goal was to predict the risk of complications for diabetic patients.
They as well introduced CBR method known as the LID (Lazy Induction of Descriptions), which was used to obtain the risk pattern of each diabetic patient. For each patient, the LID method would determine the risk of each diabetic complication according to the risk of already diagnosed patients
The main goal of DIRAS is to obtain an individual risk pattern for diabetic patients using Case-based reasoning.
DIRAS uses LID, a Case-based Reasoning method that builds a discriminant explanation of the assessed risk of complication using a heuristic based on the RLM distance.
Developed Methodology
LID Algorithm:
The LID algorithm begins with the whole set of precedents B classified into the collection of
risk classes R for a complication C, a problem p to be solved and the description D = Ø (i.e.
D has no features). D: = Ø; R = { R1 ...Rn} Function LID ( B,p,D) Sd: = Discriminatory – set (D, B) If " ei E SD Þ ei E Ri then return Ri Else, fd: = select – feature (p, B, R) D':= add-feature (fd,D) LID (SD, p, D') End if
End function
Case/risk classification:
given collection of risk classes R = {unknown, low, moderate, high, very-high}, a diabetic complication C, and a
problem p, the task of LID is to obtain the risk Ri E R of p concerning C. For each complication C, this can be
seen as a classification task where the goal is to identify the class in R to which p belongs. DIRAS solves this
classification task using LID.
if given a case base B containing patient classified into the collection of risk classes R for a diabetic complication C, and
a problem p to be classified, LID obtains the class Ri E R to which p belongs.
Intuitively, LID follows a top-down strategy to build a description D containing the most relevant features of p such
that all features in D are satisfied by a subset of cases in B. In general, cases in this subset belong to different solution
classes in R. LID adds relevant features to D until the subset of cases satisfying D belong to one unique solution class Ri.
LID takes this class Ri as the solution for the current task, i.e. Ri is the risk of p concerning C.
LID Methodology;
DIRAS constructs a symbolic explanation with the features that are
relevant in classifying a patient complication in a risk class. This symbolic
explanation is close to the justification that can be provided by an expert
for the same problem and may allow the user to focus on the critical
features for a particular patient.
Issues: The major difficulty in the diagnoses of patients with acute poisoning is usually poison identification in
the patient’s organism and circumstances of an exposure.
presented a problem of medical diagnosis which is decision support for therapy selection in case of intoxication
with drugs, psychotropes in particular
Goal:
They aimed, with their studies to find a solution to reducing the time required to come to a decision particularly
in an emergency case, as well as to compensate unavoidable lack of experience of young medical staff.
The author’s goal was to create a CBR decision support system which would be useful for the following purposes:
-- To be used by an ambulance physician in the cases of an intoxication by medicines.
-- To be used by physicians whose specialty lies outside the domain of toxicology
Case IIICase based decision support system for diagnosing intoxications by drugs by Klaus – Dieter et al
Klaus – Dieter et al described an approach for developing expert knowledge based medical
decision support system that was based on new technology of case based reasoning using
INRECA approach.
The Inreca system allows to compile some of the specific knowledge contained in the cases
into more general rules that can be efficiently evaluated and which keep the consultation time
of the system short. This approach can be viewed as integration between classical CBR and
inductive machine learning approaches.
INRECA technology approach
CBR applications in medical domains requires best match of cases and chosen the most similar cases from
the system. In medical domains, a case contains description of the symptoms observed during
examination of a patient as well as the diagnosis or the treatment that was identified by a physician.
In the toxicology domain for instance, the symptoms that are recorded in a case could appear as attributes
in the case representation, each of which has an assigned particular type, which defines the value range of
the attribute.
Thus, when a new problem is encountered, for example, a patient with an unknown intoxication must be
diagnosed; some of the symptoms must be checked and noted as a new problem case
The CBR process then proceeds by searching for the most similar known case from the case base. For this
purpose, the similarity between two cases (the problem case and the case in the case base) must be defined
through the similarity of the attributes used in the case representation (except for the target attribute).
Cases and Similarity Measures:
The Inreca approach as was proposed by the authors allows finding the most similar cases in an efficient
way. The main idea behind the approach is that the authors introduced a new indexing structure know as
the Inreca tree.
The indexing structure was based on the concept of K- d tree, a multi – dimensional binary search tree,
which was described in [2]. During cases retrieval, the tree focuses the search for the most similar cases
and thereby avoids the investigation of available cases. The search is done via a recursive tree search
procedure according to the global similarity measure SIM(X, Y). During the search, the two test procedures
‘Ball-Overlap-Bounds‘’ (BOB) and ‘Ball-Within-Bounds‘’ (BWB) were used to focus on the relevant
search region.
Approach/methods
In their system testing, the authors developed an initial CBR system for the toxicology domain, in
particular for the task of poison recognition during acute poisoning.
They selected a data set of 459 cases based on eight types of drugs for the testing. The data set was also
measured with other algorithm like the Kate CBR, Kate Induction and AS algorithm.
The result of the testing shows that the CBR approach leads to a high classification accuracy which is only
slightly worse than the accuracy of the AS algorithm which is highly optimized for the toxicology domain.
Based on their preliminary experimental result, the authors opined that a systematic approach to CBR
system development will lead to valuable case-based toxicological decision support system
Testing:
Multiple Case disorders : Medical or bank cases could be different at each point in time, at
such no matching in the case base database. This deserve special attention and treatment
because they occur very often in medical cases.
Time Constraints: Users of expert system normally do not have time to sit down and interact
with a system before they have to decide about a patients treatment.
Integration issues: A diagnostic system has to be integrated into the existing information
systems in a hospital, for instance and get from there all vital data for the patient, in most case
there are lack of proper expertise in getting the system to work together A large case base can suffer problems from storage and trade-offs.
Cases do not often include a deeper knowledge of the domain.
The Open Problems of CBR
1. Introducing a development of a CBR with graphical user interface, so as to make the CBR system
more user friendly for the expert system users. The system as well may be extended to assist the economic
analysts in the ban case in maintaining the case bases.
2. CBR System adaptation. Adaptation remains a problem in CBR use in medicine and only a few
systems are used in a clinical setting. More work on how to adapt pervious CBR system is an area for
future research.
Implementation of CBR system through Web-based electronic medical records,i.e the concept of well
patient care records (having access to and monitoring your health even if you are well), personalization,
handheld computing for clinical decision support in hospitals and the nexus of CBR and evidence-based
practice, which all argue well for using CBR in medicine are all probable future research directions.
Proposed Research Directions:
In other to overcome the problems, many expert sysytem researchers are finding means to
develop variants of CBR system to find efficient ways of building a robust CBR system in
favor of CBR components integrated into clinical information and administrative systems. In
this regard, part II of the report reviewed some research work done on CBR in bank and
medicine for effective administration and management of new case problems.
Above all, CBR is adequate for bank and medical diagnosis for number of reasons as outlined
in the report which include easy and fast administration to patients through quick retrieval of
patients past similar cases from the system to solve a new case.
Conclusion
Maria de Fatima T and Luis M.B. A case-based reasoning approach for predicting bank lending decisions. Banco de Portugal/Research Department.
Abdel-Badeeh M.S. Case based reasoning technology for medical diagnosis. World Academy of Science, Engineering and technology,2007.
Klaus D.A, Ralph B and Stefan W. Case based reasoning for medical decision support task: the INRECA approach. In: Artificial Intelligence in Medicine
Journal, Vol.12,No.1, Jan. 1998, 25-41
Agnar A. Case-Based Reasoning- An Introduction. Department of informatics,University of Trondheim, Norway.
Koton P : Using experience in learning and problem solving.Massachusetts Institute of Technology,Laboratory of Computer science, October 1989.
Alof K.D. Knowledge acquisition in the domain of CNC machine centers; the MOLTKE approach.Third European Workshop on Knowledge Based
Systems,Paris,July 1989.pp 180-195
Richter A.M and Weiss S. Similarity, uncertainty and case-based reasoning in PADEX. Proceedings of the Eighth international Joint conference on
Artificial Intelligence, IJCAI, Karlsruhe, 1991.
Kitano H. Challenges for massive parallelism. IJCA!-93, proceedings of the Thirteenth International Conference on Artificial Intelligence, Chambery,
France,1993.
References