Advanced
Context-Awareness Healthcare for Disease Reasoning Based on Fuzzy Logic
Context-Awareness Healthcare for Disease Reasoning Based on Fuzzy Logic
Journal of Electrical Engineering and Technology. 2016. Jan, 11(1): 247-256
Copyright © 2016, The Korean Institute of Electrical Engineers
This is an Open-Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
  • Received : June 23, 2015
  • Accepted : November 03, 2015
  • Published : January 01, 2016
Download
PDF
e-PUB
PubReader
PPT
Export by style
Article
Author
Metrics
Cited by
TagCloud
About the Authors
Byung-Kwan Lee
Dept. of Computer Engineering, Catholic Kwandong University, Korea. (bklee@ cku.ac.kr)
Eun-Hee Jeong
Corresponding Author: Dept. of Regional Economics, Kangwon National University, Korea. (jeongeh@kangwon.ac.kr)
Sang-Sik Lee
Dept. of Biomedical Engineering, Catholic Kwandong University, Korea. (lsskyj@cku.ac.kr)

Abstract
This paper proposes Context-Awareness Healthcare for Disease Reasoning based on Fuzzy Logic. It consists of a Fuzzy-based Context-Awareness Module (FCAM) and a Fuzzy-based Disease Reasoning Module (FDRM). The FCAM computes a Correlation coefficient and Support between a Condition attribute and a Decision attribute and generates Fuzzy rules by using just the Condition attribute whose Correlation coefficient and Support are high. According to the result of accuracy experiment using a SIPINA mining tool, those generated by Fuzzy Rule based on Correlation coefficient and Support (FRCS) and Improved C4.5 are 0.84 and 0.81 each average. That is, compared to the Improved C4.5, the FRCS reduces the number of generated rules, and improves the accuracy of rules. In addition, the FDRM can not only reason a patient’s disease accurately by using the generated Fuzzy Rules and the patient disease information but also prevent a patient’s disease beforehand.
Keywords
1. Introduction
IT technologies such as Internet, Mobile, and Ubiquitous are now changing the paradigm of medical care from provider-oriented medicine to user-oriented medicine, from treatment medicine to preventive medicine, and from disease-oriented medicine to wellbeing-oriented medicine. Thanks to them, medical customers can be diagnosed at an early stage of disease with mobile devices that are connected to computers at any time or anywhere. For this, health care providers are doing a lot of researches on individually-customized medical service to provide their customers with the best medical service.
In particular, Data Mining Technique was used to analyze the medical information of a patient’s systemic lupus erythematosus in the RX project of Stanford University in the 1980’s [1] . Since then, Data Mining technique using machine learning has been applied to many fields of medical diagnosis such as thyrotoxic myopathy [2 - 4] , rheumarthritis [5] , cardiovascular diseases [6] , Neuropsychosis [7] , etc [8] .
This paper proposes Context-Awareness Healthcare for Disease Reasoning based on Fuzzy Logic. It consists of the Fuzzy-based Context-Awareness Module (FCAM) and the Fuzzy-based Disease Reasoning Module (FDRM).
The FCAM fuzzifies disease information, manages a patient’s health condition by collecting the context-awareness information such as a patient’s physiological information, medical record, family history in real time, and generates the fuzzy rules on a disease. The FDRM reasons a disease with the Fuzzy rules and a patient’s Context-Awareness information collected in the FCAM accurately and prevents a patient from an unexpected disease beforehand.
The remainder of this paper is organized as follows. Section 2 discusses the related works on a Fuzzy Logic System and C4.5. Section 3 shows the Context-Awareness Module and Disease Reasoning based on Fuzzy Logic. Section 4 shows the estimation result and finally in section 5, we describe our conclusions.
2. Related works
- 2.1 Fuzzy logic system
“Fuzzy Logic” was introduced with the 1965 proposal of fuzzy set theory by Lotfi A. Zadeh [9] . The Fuzzy Logic has been applied to many fields, from control theory to artificial intelligence.
Specially, a Fuzzy Logic System can be defined as the nonlinear mapping of an input data set to a scalar output data [10 , 11] . A Fuzzy Logic System consists of four main parts: fuzzifier, rules, inference engine, and defuzzifier. The general architecture of a Fuzzy Logic System is shown in Fig. 1 .
PPT Slide
Lager Image
A Fuzzy logic system
The process of a Fuzzy Logic system is explained in Algorithm 1 [9] .
PPT Slide
Lager Image
- 2.1.1 Linguistic variables
Linguistic variables are the input or output variables of the Fuzzy Logic System whose values are words or sentences from natural language, instead of numerical values. For example in the Fuzzy Logic System the linguistic variables are expressed as follows according to a patient’s temperature.
temperature(t)={hypothermia, normothermia, mild fever, high fever}
- 2.1.2 Membership function
Membership functions are used in the fuzzification and defuzzification steps of the Fuzzy Logic System. Membership functions can have several different shapes such as Fig. 2 [9 - 11] .
PPT Slide
Lager Image
Membership functions
The most commonly used shapes are triangular, trapezoidal, Gaussian and bell shaped membership function.
This paper uses triangular, trapezoidal, and singleton membership function for a patient’s disease reasoning.
- 2.1.3 Fuzzification
A Fuzzy Logic system uses linguistic variables instead of numerical variables. The process of converting a numerical variable (real number or crisp variable) to a linguistic variable (fuzzy number) is called fuzzification. The simplest form of membership function is triangular membership function.
- 2.1.4 Defuzzification
The reverse fuzzification is called defuzzification. The use of a Fuzzy Logic System inference engine produces the output in a linguistic form. According to real world requirements, the linguistic variables have to be transformed to crisp output. Weighted average method is the best well-known defuzzification method for Sugeno type fuzzy controller [12 , 13] .
This paper produces a patient's disease reasoning result in this weighted average method.
- 2.2 Context-awareness
Context-awareness is defined by Brown [14] , Ryan [15] , and Dey [16] .
Brown et al. [14] defines context as location, identities of the people around a user, the time of day, season, temperature, etc. Ryan [15] defines context as a user’s location, environment, identity and time. Dey [16] enumerates context as the user’s emotional state, focus of attention, location and orientation, date and time, objects, and people in a user’s environment [17] .
This paper defines Context-Awareness as a patient’s physiological information, personal information, medical record, and family history.
In addition, it provides a patient’s disease accurately by combining Fuzzy Theory such as Fuzzy Set, Fuzzy Rule, etc with the Context-Awareness information.
- 2.3 C4.5 algorithm
C4.5 is an algorithm used to generate a decision tree developed by Ross Quinlan [18 , 19] . C4.5 is an extension of Quinlan’s earlier ID3 algorithm. The decision trees generated by C4.5 can be used for classification, and for this reason, C4.5 is often referred to as a statistical classifier [20] .
In general, the steps in C4.5 algorithm to build decision tree are [21 , 22] :
- Choose attribute for root node
- Create branch for each value of that attribute
- Split case according to branches
- Repeat process for each branch until all cases in the branch have the same class.
Choosing which attribute to be a root is based on highest gain of each attribute. To count the gain, formula (1) is used to count the gain [21 , 22] .
PPT Slide
Lager Image
Here, {S 1 , … , S i , … , S n } the partitions of S according to values of attribute A.
n is the number of attributes A.
|S i | is the number of cases in the partition S i .
|S| is the total number of cases in S.
The entropy is gotten by formula (2) [21 , 22] .
PPT Slide
Lager Image
Here, S is case set, and n is the number of cases in the partitions S. P i is the Proportion of S i to S.
3. Context-Awareness Healthcare for Disease Reasoning based on Fuzzy Logic
The “Context-Awareness Healthcare for Disease Reasoning based on Fuzzy Logic” proposed in this paper consists of a Fuzzy-based Context-Awareness Module (FCAM) and a Fuzzy-based Disease Reasoning Module (FDRM). Its total structure and data flow are shown in Fig. 3 .
PPT Slide
Lager Image
The total structure and data flow
- 3.1 Overview
The Fuzzy-based Context-Awareness Module (FCAM) fuzzifies all the disease information beforehand, generates Fuzzy Rules with the disease information and collects the Context-Awareness information such as physiological information, medical records, personal information, and family history to check a patient’s current health condition. Then, it judges a patient’s disease after comparing the fuzzified disease information with a patient’s Context-Awareness information.
In particular, when it generates Fuzzy Rules, it not only decreases the number of conditions within a rule by excluding the condition attributes whose Correlation coefficient and Support is the lowest about elements but also improves the accuracy of disease reasoning with the better accuracy of rules.
The Fuzzy-based Disease Reasoning Module (FDRM) reasons a patient’s disease accurately by using the Fuzzy Rules and the Context-Awareness information collected in the FCAM and then prevents a patient from an unexpected disease by informing them of the result.
- 3.2 Fuzzy-based context-awareness module (FCAM) design
The FCAM collects a patient’s Context-Awareness information and classifies it into the following 4 categories.
- A patient’s physiological information such as chest pain, cholesterol level, blood sugar, blood pressure, ECG, etc
- A patient’s medical records
- A patient’s personal information such as age, weight, height, smoking, drinking, job, etc.
- A patient’s family history
Besides, the FCAM defines Fuzzy Set, data range and linguistic variables about each element with the collected Context-Awareness information in Table 1 [23 , 24] .
Fuzzy set, range, and linguistic variables
PPT Slide
Lager Image
Fuzzy set, range, and linguistic variables
Fig. 4 shows the membership function and fuzzification of cholesterol [24] .
PPT Slide
Lager Image
Fuzzification and membership function
Fig. 5 shows the procedure in which the FCAM is processed.
PPT Slide
Lager Image
The procedure of FCAM
To begin with, the FCAM collects a patient’s Context-Awareness information in real time, judges the patient’s health condition by using Table1, and informs the patient of the result. At this time, if the patient’s health condition is abnormal, the FCAM transfers the patient’s family history and medical records to the FDRM so that the FDRM can reason the patient’s disease.
For example, Bob has blood pressure 150 and both of his parents suffer from high cholesterol and diabetes. When the FCAM receives such cholesterol information about Bob, it judges his health condition as “abnormal” because the “linguistic variable in Table1 is “high”. Therefore, because his health condition is abnormal, the FCAM transfers a “attention: message” to him and the information on the his medical records and family history to the FDRM.
- 3.2.1 Fuzzy rule generation
The FRCS (Fuzzy Rule based on Correlation coefficient and Support) algorithm proposed in this paper generates Fuzzy Rules by using the disease information stored in the database. At this time, the FRCS algorithm computes the Correlation coefficient and Support about a Decision Attribute and a Condition Attribute. Then, because the FRCS selects the Condition Attribute which has the highest Correlation coefficient and Support, it not only reduces the number of conditions within rules, but also improves the accuracy of rules.
shows the flow of the FRCS algorithm.
PPT Slide
Lager Image
shows the flow of the FRCS algorithm.
The following example shows that the FCAM generates Fuzzy Rules by using the FRCS algorithm and Heart Disease [24] . Table 2 shows the attributes of Heart Disease data and Fig. 6 shows its experimental data.
The attributes of Heart Disease and statistical values
PPT Slide
Lager Image
The attributes of Heart Disease and statistical values
PPT Slide
Lager Image
Experimental data (information table)
(1) The attribute selection of Fuzzy Rule
The FCAM selects Fuzzy rule’s attributes by computing the Correlation coefficient and Support of each attribute with the FRCS algorithm.
[The 1 st step] The FCAM classifies the value of a Decision attribute (a14) into no or yes and computes a Correlation coefficient and Support about each attribute.
At this time, the FCAM computes the Correlation Coefficient and Support by using the expression (3) and (4). Table 3 shows the result of the Correlation Coefficient and Support.
The Conditional support of each attribute
PPT Slide
Lager Image
The Conditional support of each attribute
PPT Slide
Lager Image
PPT Slide
Lager Image
Here, CC is Correlation Coefficient, S Support, CA Condition Attribute, DA Decision Attribute.
For example, the number of data which is a1<60 among 270 experimental data is 186. The number of data whose Decision attribute (a14) is “YES” and “NO” among the 186 is 72 and 114 each. At this time the Support of “YES” by using expression (4) is 72/(72+114) = 0.39.
[The 2 nd step] The FCAM selects as a Condition attribute of Fuzzy rules the attribute whose correlation coefficient and support is the highest. For example, because a13 meets this condition, it is selected as the condition of Fuzzy rules.
With these selected attributes, Correlation coefficient and Support are computed in the generation step of Fuzzy Rules repeatedly and only the attributes whose Correlation coefficient and Support are high are used as a Fuzzy Rule in the FCAM.
(2) Fuzzy Rule Generation
The FCAM computes the Correlation coefficient and Support about a Condition attribute with the selected attributes and decides the order of rule conditions about each Fuzzy Rules.
The FCAM generates the condition of Fuzzy rules with the attributes the FRCS algorithm is satisfied with.
[The 1 st step] The FCAM selects as the condition of Fuzzy rules the attributes whose Correlation coefficient is the highest. That is, because a13 meets this condition in the Table 3 , it is selected as the condition of Fuzzy rules
[The 2 nd step] The FCAM arranges the experimental data on the basis of the attributes selected in the 1 st step and computes the Support about each attribute with the expression (4).
For example, the FCAM extracts just the data whose attributes are a13<4. Consequently, the FCAM extracts 152 data and computes the Correlation Coefficient and Support with expression (3) and (4). The Table 4 shows the result of the 2 nd step.
The result of correlation coefficient and support
PPT Slide
Lager Image
The result of correlation coefficient and support
[The 3 rd step] The FCAM selects as Fuzzy rules the attributes whose Support in the 2 nd step is highest. That is, the FCAM selects as the condition of Fuzzy rules the a3 whose Support is 0.90.
Fig. 7 shows an example of the Fuzzy Rules generated by the FCAM. Its Fuzzy Rules in Fig. 7 can have several rule conditions and each rule condition can be called the 1 st rule condition, the 2 nd rule condition, etc according to each location.
PPT Slide
Lager Image
The example of fuzzy rules
[The 4 th step] The FCAM arranges the experimental data on the basis of the a3 selected in the 3 rd step and decides the conditions of Fuzzy rules by repeating the 1 st step ~the 3 rd step. At this time, the FCAM selects the conditions of Fuzzy whose Support is less than 0.35.
[The 5 th step] After the FCAM selects the rule condition of each Fuzzy Rule in this way, it generates a Fuzzy Rule and stores it in the database. Because the Fuzzy Rule generated by the FRCS algorithm is selected by computing the Conditional Support among each attribute, their accuracy is greater than that generated by the Improved C4.5 algorithm.
The rules generated by the Improved C4.5 algorithm and the FRCS algorithm are summarized in Table 5 .
The rules of the improved C4.5 and the FRCS algorithm
PPT Slide
Lager Image
The rules of the improved C4.5 and the FRCS algorithm
- 3.3 Fuzzy-based disease reasoning module (FDRM) design
The Fuzzy-based Disease Reasoning Module (FDRM) reasons a patient’s disease by using the Fuzzy Set, Context-Awareness and Fuzzy Rules. Fig. 8 shows the total flowchart of the FDRM.
PPT Slide
Lager Image
The flowchart of the FDRM
[The 1 st step] The FDRM applies the Fuzzy Set, range, linguistic variables and the Fuzzy Rules to Fuzzy Logic.
[The 2 nd step] The Fuzzy Logic generates a membership function with Fuzzy set, range, and linguistic variables.
[The 3 rd step] The FDRM assigns a patient’s Context-Awareness information to the membership function and converts the result to Fuzzy value.
For example, in case the ST depression and blood pressure of Bob is 2.1311 and 150 each, he has the Fuzzy value of Fig. 10(a) and (b) .
PPT Slide
Lager Image
The result after processing FDRM
[The 4 th step] The FDRM estimates the Fuzzy Rule of Fig. 9 and prints out the result value like Fig. 10(c) .
PPT Slide
Lager Image
Fuzzy Rule (ST depression and Blood pressure)
[The 5 th step] The FDRM computes a disease reasoning value about a patient’s Context-Awareness information by using the defuzzification formulae [12] of Table 6 .
Defuzzification
PPT Slide
Lager Image
Defuzzification
[The 6 th step] The FDRM notifies patients of the health information according to the computed result value. If the patient’s health condition is abnormal, it prevents a patient from an unexpected disease beforehand by transferring the information to him rapidly.
4. Analysis
The FCAM and FDRM proposed in this paper are estimated based on the accuracy of Fuzzy Rules and Disease Reasoning by using the experimental data of Heart Disease [25] .
- 4.1 The accuracy estimation for fuzzy rules
The Heart Disease consists of 14 attributes and the explanation about each attribute is shown in the Table 2 . This experimental data is made up of total 270 data and the Fuzzy Rules are generated with the 270.
Consequently, as shown in the Table 7 , the number of the decision rules generated by the FRCS and the Improved C4.5 algorithm each is 6 and 7. This paper estimates the accuracy in comparison with the rules generated by the Improved C4.5 and the FRCS algorithm using SIPINA mining tool [26] . The Table 7 shows the estimated result. The number of the attributes used in the FRCS is smaller than that of the Improved C4.5, but the FRCS algorithm is more accurate. Therefore, it is concluded that the FRCS is more excellent than the Improved C4.5.
The comparison of the FRCS and the Improved C4.5 algorithm
PPT Slide
Lager Image
The comparison of the FRCS and the Improved C4.5 algorithm
Fig. 11 shows the rule accuracy between Improved C4.5 and FRCS. The accuracy of rules is better if the gap in the graph is narrower. Therefore, Fig. 11 (b) shows that the rule accuracy of FRCS is the higher.
PPT Slide
Lager Image
The comparison of their accuracy
- 4.2 The accuracy estimation for disease reasoning
Table 8 shows that the rules generated by the FRCR algorithm were converted to the linguistic variables of Fuzzy Set. The rules made out of family history and medical records were added to the Table 8 .
Fuzzy reasoning rules
PPT Slide
Lager Image
Fuzzy reasoning rules
Table 9 shows the Fuzzy Set on heart disease as a result of a disease reasoning rule, and Fig. 12 shows the fuzzification of heart disease.
Fuzzy set of heart disease
PPT Slide
Lager Image
Fuzzy set of heart disease
Experimental data and reasoning result
PPT Slide
Lager Image
*a: absent, p: present
PPT Slide
Lager Image
Fuzzification of Heart_disease
The Accuracy of Disease Reasoning was estimated with the randomly extracted 10 values of Heart Disease data.
The result shows that the accuracy of Disease Reasoning in the FDRM is 80%. The accuracy estimation is done by applying the randomly selected values of experimental data [25] to the Fuzzy rules of Table 8 . The accuracy is computed by using expression (5)
PPT Slide
Lager Image
where, the number of data means the result which is true in comparison.
Therefore, when the number of experimental data is 200, the accuracy of Disease Reasoning in the FDRM is 86%.
As shown in Fig. 13 , as the number of experimental data gets increased, the accuracy of Reasoning gets better.
PPT Slide
Lager Image
The accuracy estimation according to the number of experimental data
5. Conclusion
This paper proposes Context-Awareness Healthcare for Disease Reasoning based on Fuzzy Logic. It consists of the Fuzzy-based Context-Awareness Module (FCAM) and the Fuzzy-based Disease Reasoning Module (FDRM).
The FCAM collects the Context-Awareness information such as physiological information, medical records, family history, etc and manages a patient’s disease and the FDRM prevents a patient’s disease by reasoning his disease with the information of the FCAM.
The FCAM and the FDRM have the following characteristics.
First, the FCAM manages patient’s diseases by rapidly notifying a patient of his health condition after fuzzifying his collected Context-Awareness information.
Second, the FCAM computes the Correlation coefficient and Support between Condition attribute and Decision attribute. The FRCR algorithm generates Fuzzy Rules by using the attributes whose Correlation coefficient and Support are high and strengthens the relation between attributes, contrary to the Improved C4.5 algorithm. Therefore, the FRCR algorithm decreases the number of attributes used for Fuzzy Rules more than the Improved C4.5 algorithm and improves the accuracy of rules more.
Third, the FDRM reasons a patient’s disease by using disease information, Fuzzy Set, Fuzzy Rule and predicts a patient’s disease by the reasoning result
Acknowledgements
This work was supported by research fund of Catholic Kwandong University (CKURF-201407130001).
BIO
Byung-Kwan Lee He received his B.S. degree from Pusan National University in 1979, the M.S. degree in Computer Science from Chung-Ang University in 1986 and the ph.D. degree in Computer Science from Chung-Ang University in 1990 in Korea. He has been a professor of Computer Science at Catholic Kwandong University in Korea since 1988. He was a visiting professor at Saginaw Valley State university, Michigan, USA during 2000~2001. He is a permanent member of the KISS and KIPS. His current research interests are network security, Wireless Sensor Network Security, Internet of Things and Big Data.
Eun-Hee Jeong She received her B.S. degree from Kangneung National University in 1991, the M.S. degree in Computer Science from Kwandong University in 1998 and the Ph.D. degree in Computer Science from Kwandong University in 2003 in Korea. She has been a professor of department of Regional Economics at Kangwon National University in Korea since 2003, Sept. She is a regular member of the KSII. Her current research interests are Sensor Network, IT security, web programming, and e-commerce security.
Sang-Sik Lee He was born in Sancheong, Korea, in 1968. In 1994 he graduated in Bio-mechatronic engineering at the Sungkyunkwan University, Korea. He received his master degree in Engineering (Mechatronics system) in 1996 and his Ph.D in 2000 from Sungkyunkwan University, Korea. He is currently an assistant professor at the college of Engineering at Catholic Kwandong University in Gangneung, Korea. His main research interests are in the area of Biomedical engineering and u-Health systems.
References
Blum R. L. 1981 “Displaying clinical data from a time-oriented database,” Computers in Biology and Medicine 11 (4) 197 - 210    DOI : 10.1016/S0010-4825(81)80022-4
KI Hojker S. , Jauk A. , Fidler V. , Porenta M. 1988 “Expert system’s development in the management of thyroid diseases,” European Congress for Nuclear Medicine Milano
Horn W. 2001 “AI in medicine on its way from knowledge-intensive to data-intensive systems,” Artificial Intelligence in Medicine 23 (1) 5 - 12    DOI : 10.1016/S0933-3657(01)00072-0
Quinlan R. CP , Horn KA , Lazarus L. 1987 Applications of expert systems Addison-Wesley Boston “Inductive knowledge acquisition: a case study. In: JR Q.,” 137 - 156
Zupan B. , Dzeroski S. 1998 “Acquiring background knowledge for machine learning using function decomposition: a case study in rheumatology,” Artificial Intelligence Medicine 14 (1-2) 101 - 117    DOI : 10.1016/S0933-3657(98)00018-9
Cohen ME , Hudson DL 2006 “Neural network models for biosignal analysis,” Conference Proceeding IEEE Engineering in Medicine Biology Society 1 3537 - 3540
Chun FK , Karakiewicz PI , Briganti A. , Walz J , Kattan MW , Huland H , Graefen M. 2007 “A critical appraisal of logistic regression-based nomograms, artificial neural networks, classification and regression-tree models, lookup tables and risk-group stratificationmodels for prostate cancer,” BJU International 99 (4) 794 - 800    DOI : 10.1111/j.1464-410X.2006.06694.x
Lee SM , Park RW 2009 “Basic Concepts and Principles of Data Mining in Clinical Practice,” Journal of Korea Society Medical Informatics 15 (2) 175 - 189    DOI : 10.4258/jksmi.2009.15.2.175
Zadeh LA 1965 “Fuzzy set,” Information and Control 8 338 - 353    DOI : 10.1016/S0019-9958(65)90241-X
A Short Fuzzy Logic Tutorial Available at
Mendel J. 1995 “Fuzzy logic systems for engineering: a tutorial,” proceedings of the IEEE 83 (3) 345 - 377    DOI : 10.1109/5.364485
1997 Defuzzification formular Fuzzy control programming, Technical report International Electrotechnical Commision
Dhote Nitin K. , Helonde J. B. 2014 “Improvement in Transformer Diagnosis by DGA using Fuzzy Logic” J Electr Eng Technol 9 (2) 615 - 621    DOI : 10.5370/JEET.2014.9.2.615
Brown PJ , Bovey JD , Chen X. 1997 “Context-Aware Applications: From the Laboratory to the Market-place, ” IEEE Personal Communications 4 (5) 58 - 64    DOI : 10.1109/98.626984
Ryan N. , Pascoe J. , Morse D. , Gaffney V. , van Leusen M. , Exxon S. 1997 Computer Applications in Archaeology “Enhanced Reality Fieldwork: the Context-Aware Archaeological Assistant,”
Dey AK 1998 “Context-Aware Computing: The Cyber-Desk Project,” AAAI 1998 Spring Symposium on Intelligent Environments Technical Report SS-98-02 51 - 54
Anind KD , Gregory DA 1999 “Towards a Better Understanding of Context and Context- Awareness,” Available at
Quinlan JR 1993 “C4.5: Programs for Machine Learning,” Morgan Kaufmann Pulishers Inc San Francisco
Quinlan JR 1996 “Learning decision tree classifiers,” ACM Computing Surveys (CSUR) 28 (1) 71 - 72
http://en.wikipedia.org/
Craw S “Case Based Reasoning: Lecture 3: CBR Case-Base Indexing,” Available at staff/smc/teaching/cm3016/Lecture-3-cbr-indexing.ppt
Kusrini , Hartati Sri 2007 “Implementation of C4.5 Algorithm to evaluate the cancellation possibility of new student applications at stmik amikom yogyakarta,” Proceedings of the International Conference on Electrical Engineering and Informatics Institut Teknologi Bandung Indonesia June 17-19 623 - 626
Parvin Rehana , Abhari Abdolreza “Fuzzy database for Heart Disease Diagnosis,” available at July.15 2014
Adeli Ali , Neshat Mehdi “A Fuzzy Expert System for Heart Disease Diagnosis,” Proceedings of the International MultiConference of Engineers and Computer Scientists 2010 (IMECS 2010) Hong Kong March 17-19, 2010 1
Heart disease http://archive.ics.uci.edu/
SIPINA mining tool http://eric.univ-lyon2.fr/~ricco/sipina.html