Case-based Reasoning for Skin Diseases Diagnose
using Minkowski Distance
Mihuandayani
1
, Yufika Sari Bagi
1
and Theofani Christi Irene Momongan
1
1
Department of Information System, STMIK Multicom Bolaang Mongondow, Kotamobagu, Indonesia
Keywords: Case-based Reasoning, Skin Disease, Minkowski Distance.
Abstract: The skin is one of the crucial organs in the human body because it functions to receive stimuli such as touch,
pain and other influences from the outside. According to the data, skin disease is the third out of the ten most
diseases in Indonesia, and the skin specialists carry out the treatment of patients. However, due to the
limitations of the experts resulting in the slow handling of patients so that needed a tool that can help diagnose
patients with skin diseases. Case-Based Reasoning is one of the problem-solving techniques to build a system
by making decisions from new cases based on the solutions of the old cases that are closest to the new case.
The process of diagnosis is to enter new problems, compared to the old case, and then calculate the value of
proximity using Minkowski distance. This study produced a case-based reasoning system to diagnose skin
diseases due to the virus and bacterial infections based on selecting the symptoms suffered by patients and
providing treatment solutions. The testing is done by comparing the expert diagnosis data and system
diagnostic results with 92.10 % accuracy.
1 INTRODUCTION
The skin is a vital organ in humans located in the
outer layer of the body, which functions to receive
stimuli from the outside. Human skin condition based
on data (Karimkhani et al., 2017) in 2013, contributed
around 1.79% of the world disease burden. Damage
to the skin barrier is one of the common problems
related to nursing. Dermatological diseases can affect
significantly in physical, psychological and social
(Hay et al., 2014). Unhealthy skin can cause various
skin diseases, so it needs to maintain skin health
earlier to avoid diseases. A person’s body skin that is
affected by the disease can interfere with the
appearance and activities of the person. Skin disease
is often underestimated because it tends to be
harmless and not cause death. The 2010 Indonesian
Health Profile data shows that skin disease is ranked
as the third out of the ten most diseases in outpatients
in Indonesian hospitals. The incidence of skin
diseases in Indonesia is still relatively high and
becomes a significant problem. Various skin diseases
can be caused by several factors, such as the
environment and bad daily habits, climate change,
viruses, bacteria, fungi, allergies, endurance, and
others.
Skin diseases due to virus infections are classified
into several types, including Verruca Vulgaris,
Herpes Zoster, Herpes Genital, and Varicella.
Verruca Vulgaris clinically in the form of solid
papules/plaque and the surface is in the form of small
papules measuring 1-3 mm, and the most frequent of
Human Papilloma Virus (HPV) infection. Herpes
Zoster is a disease caused by the Varicella-Zoster
virus, mainly affecting adults with the characteristics
of radicular, unilateral pain and hordes of vesicles
scattered according to the dermatome that is
conserved by a sensory nerve ganglion. Herpes
Genital becomes an infection of the skin disease
caused by the Herpes Simplex Virus (HSV)
transmitted through sexual contact. Varicella is a skin
disease with scattered vesicles, primarily affecting
children, which is easily transmitted by the Varicella-
Zoster virus.
Then, for skin diseases due to bacterial infections
classified into several types including Acne vulgaris
is a chronic inflammation of the pilosebaceous
follicles which is characterized by blackheads,
papules, cysts and pustules in predilection areas.
Furuncle is an acute infection of one hair follicle,
which usually experiences necrosis caused by
Staphylococcus aureus. Leprosy is a disease caused
by Mycobacterium leprae and belongs to a chronic
Mihuandayani, ., Sari Bagi, Y. and Irene Momongan, T.
Case-based Reasoning for Skin Diseases Diagnose using Minkowski Distance.
DOI: 10.5220/0010621900002967
In Proceedings of the 4th International Conference of Vocational Higher Education (ICVHE 2019) - Empowering Human Capital Towards Sustainable 4.0 Industry, pages 195-202
ISBN: 978-989-758-530-2; ISSN: 2184-9870
Copyright
c
2021 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
195
disease that attacks the peripheral nerves, skin, and
other body parts (Job, 1994). Impetigo is a superficial
and infectious pyogenic skin disease caused by
Staphylococcus or Streptococcus.
To diagnose patients suffering from skin diseases
can be known from the symptoms that appear or
experience by the patients. Handling in patients with
skin diseases is carried out by experts, namely skin
specialists, but due to limited expert expertise
resulting in slow handling of patients, so we need a
tool that can help diagnose patients with skin
diseases. Cases stored in medical records regarding
the diagnosis of skin diseases by experts to determine
the type of skin disease suffered by patients can be
reused as a reference to determine the type of skin
disease when there are new cases. Utilization of cases
that have occurred before or old cases is generally
known by the term Case-Based Reasoning (CBR).
CBR is a psychological theory of human
cognition that overcomes problems in memory,
learning, planning, and problem-solving (Slade,
1991). CBR is a computer reasoning method that
utilizes old knowledge to solve new problems.
Ancient knowledge in the form of documentation of
problems that already have solutions. The solution
can be used to solve similar new problems. The CBR
method was chosen because of its advantages, namely
the problem-solving process using existing cases and
having a revision process used to correct diagnostic
errors or inaccurate solutions, the results of the
revision can be stored as a new knowledge base on
the system, so that the system can continue to
develop. Then, distance metrics are widely used in the
estimation of similarity. In this study, the method of
calculating distances between cases uses Minkowski
distance. Minkowski distance is a metric in vector
space where the n-1 distance is also called the
distance of a city block and is often considered a
generalization of two distance, the Euclidean distance
and the distance of Manhattan (John, 1995). Besides,
in the study of Distance (Sreedevi and
Padmavathamma, 2015) and also (Karakoc et al.,
2006) explained that the Minkowski distance method
is relatively better than other distance methods.
The system that was built only covered a few
types of skin diseases, namely skin diseases due to
viruses and bacteria in adults, then the process of
revising the results of the diagnosis manually by
experts. The testing process was carried out to
measure the performance of Minkowski Distance
with data based on medical records for five years of
work, obtained from the Clinic dr. Giana Sugeha,
Sp.KK. Based on the background problem, this study
produced a system to solve diagnosis in cases of skin
diseases using CBR and the Minkowski distance
calculation method. This research can help health
workers to diagnose patients with the virus and
bacterial skin diseases, which can be used as a
reference before providing treatment therapy
recommendations.
2 RELATED WORKS
Research that applies CBR to cases of skin disease
has been done before. While in this study CBR to
diagnose skin diseases due to viruses and bacteria
with several types of diseases such as Verruca
Vulgaris, Herpes, Varicella, Furuncle, Leprosy, Acne
Vulgaris, Impetigo using the Minkowski distance
calculation method. By entering the symptoms
suffered by the patient will produce the final result
that is the illness.
Other studies related to the computational
efficiency of the k-means algorithm with distance
metrics find similar data objects that lead to the
development of robust algorithms for data mining
functionality. The research experiment applied the K-
means algorithm with three metrics, namely
Euclidean distance, Manhattan distance and
Minkowski distance. The results obtained that the
selection of distance metrics has a vital role in
clustering (Singh et al., 2013).
Other studies discuss the development of
Minkowski distance generalizations using IOWA,
and it is referred to as the Minkowski OWA operator
(IMOWAD) or induced long-distance OWA operator
(IGOWAD) by obtaining various distances.
Development of new application approaches in the
problem of decision making about investment
selection (José and Montserrat, 2011).
Research that discusses early detection of diabetes
that helps prevention measures using genetic
algorithms with the Multimodal Evolution algorithm.
The research that discusses the diagnosis of diabetes
is carried out using Genetic Algorithm and distance
calculations, including the distance of Minkowski on
the PIMA Indian dataset. Through the calculation of
the distance can be proven better accuracy using
Minkowski Distance (Sreedevi and Padmavathamma,
2015).
The research discussion is related to image
processing of various types of skin cancer by
proposing a method for detecting two types of skin,
one is cancerous skin, and the other is affected but not
cancerous skin based on several feature values. Some
values extracted from the Gray Level Co-occurrence
Matrix (GLCM) also include in Euclidean distance,
ICVHE 2019 - The International Conference of Vocational Higher Education (ICVHE) “Empowering Human Capital Towards Sustainable
4.0 Industry”
196
Manhattan Distance, Minkowski Distance, and
Hamming Distance. This process is a secure system
rather than a doctor’s biopsy procedure. The system
consumes less time and gets better results than
conventional systems. This study shows that the
combination of co-occurrence of matrix and neural
network provides a technique for detecting cancer
cells and non-cancer cells (Bhuiyan et al., 2015).
Research analyzing and comparing images based
on twelve distance calculations including the distance
of Minkowski, Euclidean, Mahalanobis, Manhattan,
Chebychev, and others for diagnosis of 176 images of
dermoscopic skin lesions. The result shows that
CBDIR performance can be significantly improved
by using Canberra and Bray-Curtis Distance
compared to conventional measures (Khadidja and
Mostefai, 2015).
Research that develops a theoretical basis for the
analysis and construction of shape sizes of time series
association. Some general methods of construction of
such sizes are suitable for measuring the time of the
equation of the series form and proposed form
association. Size associations of time series forms
based on Minkowski distance and data
standardization methods were considered. The cosine
similarity and Pearson correlation coefficient were
obtained as a part of the case from the proposed
general method which can also be used for the
construction of new association steps in data analysis
(Batyrshin, 2013).
Research that discusses the search for structural
similarities between small molecules focuses on the
method of the closest K-neighbor. The research
shows an optimal computation with weighted
Minkowski distance to maximize discrimination
between active and inactive compounds, then shows
the structure of KNN-based pruning data for the
distance of wLp which minimizes the time for finding
similarities. The result of the experiment shows that
the classification of KNN with Minkowski wL1
distance gets better accuracy than LDA and MLR
(Karakoc et al., 2006).
The discussion presents the Minkowski type
distance, including the Hamming, Euclidean, and
Chebyshev distances, for the fuzzy orthopair r-rung
set. It introduced Minkowski-type distance for
orthopair values, based on orthopairs rankings. Then
propose some distance in the fuzzy orthopair set of r-
rung and discuss it for multi-distribution decision-
making problems. Each element was expressed as an
ordered value pair. The first showed support for
membership and the last supports membership. The
study presents a method based on Minkowski
distance for a discussion of its application in multi-
attribute decision making (Du, 2018).
In other studies, the weighted geographical
regression model GWR was adapted to benefit from
a variety of distance metrics, where it was shown that
a well-chosen distance metric could improve the
performance of the model. Minkowski’s approach is
proposed, which allows the selection of optimal
distance metrics for a given GWR model. This
approach was evaluated in a simulation experiment
consisting of three scenarios. This result suggests that
the Minkowski approach can be useful when there is
no knowledge, understanding or insight into the ‘true’
distance metric. There is a significant drawback to the
Minkowski approach, where it is often difficult to
describe how certain Minkowski distance functions
can be measured, except for some general cases, such
as Manhattan (p = 1) or Euclidean (p = 2). This
deficiency tends to make the Minkowski approach
more suitable for predictive purposes with GWR than
for exploration or inferential purposes with GWR
(Binbin, 2015).
Research that develops new decision-making
models using induction ordered weighted average
operators and Minkowski distance fuzzy linguistic
variables. Several examples of the new method obtain
the results. Generalization of an IOWA operator that
uses order trigger variables is carried out to assess the
complex rearrangement process, blurred linguistic
information and fuzzy linguistic Minkowski distance
(Xian et al., 2014).
Research that discusses how to measure
perceptual similarities between two objects using
Minkowski type metrics. In the Minkowski metric
there is no substantial similarity to the same object,
through mining a broad set of visual data, the study
has found the perception of the function of distance
by calling a function that is found to be a partial
dynamic function. When compared empirically with
dynamic partial functions, Minkowski type distance
functions in shooting and image capture transition
detection, DPF performs significantly better (Li et al.,
2003). Research that presents cellular-based medical
assistance aimed at diagnosing skin diseases using
CBR and image processing to increase awareness of
disease prevention. CBR is used to determine the
symptoms of skin diseases based on image data as
setting a new knowledge base (Aruta et al., 2015).
In a study comparing 14 distance measurements,
including Minkowski Distance and modification
between feature vectors with the performance of the
primary component, PCA method proposed a
modification of sum square error (SSE) -based
distance. The experiments showed that the proposed
Case-based Reasoning for Skin Diseases Diagnose using Minkowski Distance
197
distance measure is the first three best steps for the
characteristics of different biometric systems.
Besides, it shows that using an algorithmic
combination of distance measurements gets better
results than using distances separately (Perlibakas,
2004). There are various uses of distance calculation
and its advantages in previous studies. This study
proposes calculations using Minkowski Distance in
the case of skin disease caused by viruses and bacteria
to be able to diagnose the skin disease and produce a
system that can help patients who have symptoms of
skin disease.
3 RESEARCH METHODOLOGY
This study has several stages or methods used,
ranging from data collection to calculation of results
for concluding. In the initial stages, observations are
made on the condition and procedures of a clinic. At
present, the process that occurs is that the patient
registers at the clinic then the nurse records the
patient’s medical record information from the
examination of the patient’s condition related to
symptoms, and blood pressure. The patient waits
according to the queue number. Next, the nurse is
taken to the doctor’s office. The doctor will make a
diagnosis, and further examination then gives a
prescription by the illness. Furthermore, the patient
will complete the stages to get the medicine according
to the doctor’s prescription.
The next stage is the data collection is done by
conducting interviews directly to those who have the
capacity and information needed in this research,
namely dr. Giana Sugeha, Sp.KK, as an expert in skin
and genital specialist. Furthermore, literature studies
are carried out on textbooks, research journals, and
other references. Documentation was carried out for
taking medical records of patients on sample cases
related to skin diseases caused by viruses and bacteria
for five years of work obtained from the Clinic dr.
Giana Sugeha, Sp.KK.
Data analysis was obtained to determine data
processing needs, especially in the application of
CBR, which is a problem-solving method that uses
knowledge of previous experience to solve new
problems (Pal and Shiu, 2004). A problem is solved
by looking for the same old case, if found, then the
solution of the two is also identical. However, if it is
not found, the system looks for old cases that have the
highest similarity and needs to be adapted so that the
problem founds a solution. This solution is done
based on similar situations, referred to as cases, which
have previously been stored in the system.
The stages of the CBR process are new problems
that are matched with cases that are in the case store
database and find one or more similar cases (retrieve).
Solutions that are recommended through case
matching are then reused for similar cases, the
solutions offered may be changed and adopted
(revised). If new cases do not match in the case
storage database, CBR will store the new cases
(retain) in the knowledge base. The Retrieval process
in CBR is based on the hypothesis that the solution to
previous problems can help resolve current problems,
as long as there are similarities between them. The
following is an overview of the process flow of
implementing CBR in cases of diagnosing skin
diseases shown in Figure 1.
ASSIGN
INDEXES
RETRIEVE
MODIFY
TEST
EXPLAIN REPAIR
STORE
ASSIGN
INDEXES
Indexed
Rules
Case
Memory
Repair
Rules
Similarity
Metrics
Modification
Rules
Input
Input+Indexes
PriorSolution
ProposedSolution
FailureDescription
Causal
Analysis
PredictiveFeatures
NewSolution
Working
Solution
Figure 1: CBR Processing Pipeline.
The similarity of the two points can be calculated
by distance. The more similar the two points, the
smaller the distance and vice versa. The more
different two points, the higher the distance. The
Minkowski distance calculation is used to calculate
the similarity. Minkowski distance is a metric in
Euclidean space which is a generalization of
Euclidean’s distance and Manhattan’s distance. The
following is the Minkowski distance formula.
(1)
ICVHE 2019 - The International Conference of Vocational Higher Education (ICVHE) “Empowering Human Capital Towards Sustainable
4.0 Industry”
198
In the formula, it is explained that when p = 2, the
distance is called the Euclidean distance. Then, when
p = 1, it becomes the distance of the city block.
Furthermore, the Chebyshev distance is one of the
Minkowski distance variants where p = at the
threshold.
Furthermore, to design a system that is capable of
diagnosing, modelling is needed in the form of use
case diagrams. System analysis is intended to
determine the functional and non-functional
requirements of the newly designed system.
Functional requirements related to functions must be
provided by the system to meet the primary needs and
supporting needs in running the system. Analysis of
the functional requirements of the system is modelled
using a use case diagram can be seen in Figure 2.
Figure 2: Use Case Diagram of System.
Based on the use case diagram, the nurse can input
the old case, make a diagnosis and be able to see the
diagnosis, see the basis of the case and change the
basis of the case. Whereas the doctor can diagnose
and can see the results of the diagnosis, can see the
basis of the case and make revisions in cases where
the value of the symptoms does not match the
threshold. In processing the system, the user enters
symptoms to diagnose the disease, then the system
takes the value of the selected symptoms and
compares the new case with the old case.
After that, the system calculates the proximity
between cases, namely the value and takes the case
with the lowest distance value then adjusts to the
threshold. If it matches the threshold, the system will
display the diagnostic results, namely the case code,
disease, and solution. If it does not match the
threshold, the system will save the case for revision.
Next, to find out the data processed in the design of
this study, the following is shown the relation table in
Figure. 3.
case_base
*kd_case:varchar(4)
disease:varchar(25)
solution:varchar(150)
detail_revise
*id_revise:varchar(4)
*kd_symtom:varchar(4)
value:varchar(4)
detail_case
*kd_case:varchar(4)
*kd_symptom:varchar(4)
value_symptom:Int(1)
symptom
*kd_symptom:varchar(4)
nm_symptom:varchar(55)
revise_case
*id_revise:varchar(4)
disease:varchar(25)
solution:varchar(150)
user
*id_user:Int(3)
username:varchar(15)
password:Int(8)
access:enum
‐memberName
Figure 3: Data Table Relations.
The relation table above describes the relationship
between data processing related to the system. Then,
a system that can diagnose skin diseases and manage
the similarity calculation with Minkowski Distance is
done so that the data can be evaluated in the system
test and compared with data from the clinic of dr.
Giana Sugeha, Sp.KK.
4 RESULT AND DISCUSSION
4.1 Algorithm Calculation
Data samples used in the system of diagnosing skin
diseases due to the virus and bacterial infections were
as many as 103 case data taken from the patient’s
medical record in the skin and genital specialist clinic
of dr. Giana Sugeha, Sp.KK. The data were taken in
the form of medical records of patients who have skin
diseases due to the virus and bacterial infections. The
data is then collected into case data with various types
of diseases that have been diagnosed by doctors and
solutions or therapies given. The data is divided into
65 training data and 38 testing data. Samples of the
old case data are shown in Table 1.
Case-based Reasoning for Skin Diseases Diagnose using Minkowski Distance
199
Table 1: Samples of Old Case Base Data.
Case
Number
Gender
Age
(Year)
Symptoms Disease
Diagnosis
Solution
1 M 19 Calluses
on the feet
± 3 years
Verruca
Vulgaris
Zoter 3x1
Cefat 3x
500
Nifural
3x 500
2 M 30 Warts on
hands &
feet
Verruca
Vulgaris
Ensube
3x
Futaderm
3 M 18 Warts on
hands ± 1
year
Verruca
Vulgaris
TCA
50%
Futaderm
Herclov
3x1
4 F 25 Appears
flesh on the
left leg
Verruca
Vulgaris
H Covter
TCA
5 M 28 Warts on
the feet ± 7
months
Verruca
Vulgaris
HF
Ensube
3x1
6 M 69 Warts on
the fingers
Verruca
Vulgaris
Trogge
3x1
Tropila
3x1
7 M 47 Warts in
the eyes ±
2 months,
pain,
itching
Verruca
Vulgaris
H Covter
Futaderm
Impesta
3x1
8 F 17 Calluses
on foot ± 2
months
Verruca
Vulgaris
HF
Cefar
TCA
Zoter
Ceidnan
9 F 16 Calluses
on fingers
± 2 yea
r
Verruca
Vulgaris
H Covter
TCA
10 F 64 Appears
flesh near
the eyes ±
2 months
Verruca
Vulgaris
Pibakin
The Minkowski distance calculation is used as in
formula to find out the proximity between new cases
and old cases (1), for example:
X : declared a new case
K1, K2, ……n : declared an old case
G01, G02, ….n : symptoms that represent each
case X and old cases or
previous cases such as K1,
K2, ... ..n.
0 : it is stated that the patient
does not have these symptoms
1 : it is stated that the patient has
these symptoms
Furthermore, calculating the proximity distance
between new cases and old cases is the basis of cases.
The process of calculating proximity between cases
explained to the following example:
The proximity of case X and case of K001
Value of case X:
[1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
Value of case K001:
[1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
After that, add the case of X with K001
|1-1|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-
0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-
0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-
0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-
0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-
0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-
0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-
0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-
0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-
0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-
0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3+|0-0|3 = 0 =
0
0
The Minkowski distance is derived from the
calculation of the proximity of the case of K001 to
K010. The calculation showed that the smallest
proximity level is 0. This calculation is done on all
case-based data. Thus the Minkowski distance from
the calculation of proximity determination can be
seen in the example following Table 2.
Table 2: Samples of Distance Calculation Results.
Between Distances Value
X to K001 0
X to K002 1.44
X to K003 1.25
X to K004 1.25
X to K005 1.25
X to K006 1.25
X to K007 1.25
X to K008 1.58
X to K009 1.25
X to K010 1.25
ICVHE 2019 - The International Conference of Vocational Higher Education (ICVHE) “Empowering Human Capital Towards Sustainable
4.0 Industry”
200
4.2 Evaluation
The result of system testing aims to determine the
accuracy of the system in diagnosing the diseases,
also to test the system work process whether it is
following the design that has been made using a
threshold with a predetermined value. If the threshold
value is set at 0.80, and the diagnosis is obtained more
than the threshold value of 0.80, it is necessary to
revise it. Whereas if the results obtained are the same
or less than the threshold value, then there is no need
for a revision and it can be determined that the patient
has A disease. Two tests have been compared to the
results of manual calculations with expert calculation
results. This test is conducted to find out whether the
manual calculation has a right level of accuracy by
the results of expert calculations or not. At this phase,
system testing will be carried out to check the
accuracy of the system produced whether the system
can be run according to specific standards. The
accuracy calculations used the following formula.
𝐴𝑐𝑐𝑢𝑟𝑎𝑐𝑦
  
   
𝑥 100 %
(2)
The following test results on 38 test data using
predetermined threshold values are shown in Table 3.
Table 3: Testing Using Threshold Values.
Threshold
Correct
Test
Data
Amount
of Test
Data
Accuracy
Threshold 1 32 38 84.21 %
Threshold 1.5 35 38 92.10 %
Threshold 2 35 38 92.10 %
The results of the accuracy calculation are
represented in the diagram, which is shown in Figure
4.
Figure 4: Testing Using Threshold 1, Threshold 1.5, and
Threshold 2.
Based on Table 3 and Figure. 4, it was obtained
that the calculation with the highest accuracy value
for diagnoses of skin diseases due to viruses and
infections was obtained from the threshold 1.5 and
threshold 2 with 92.10% accuracy.
5 CONCLUSIONS
Based on the research and results of system testing, it
can be concluded that this study produced a CBR
system for the diagnosis of skin and genital diseases
due to the virus and bacterial infections by calculating
the proximity between new problems and old cases
based on symptoms by accommodating case feature
values and confidence levels. The system can
diagnose diseases based on symptoms and display the
results of patient diagnosis to provide treatment
solutions. The system provides disease diagnosis
based on the similarity between old cases and new
cases. The diagnosis is considered correct if the
distance value is 1.5. The test results on skin and
genital disease test data due to the virus and bacterial
infections indicate that the system can recognize skin
and genital diseases using the Minkowski distance
method correctly with an accuracy rate of 92.10%.
ACKNOWLEDGEMENTS
STMIK Multicom Bolaang Mongondow for the
support in providing the facility which is needed
along with the study. Especially for dr. Giana Sugeha,
Sp.KK. She has given the clinic data and
documentation for the research. This good
cooperation among all sector to provide a paper that
hopes can be useful for others.
REFERENCES
Aruta CL, Calaguas CR, Gameng JK, Prudention MV, and
Lubaton AACJ. Mobile-based Medical Assistance for
Diagnosing Different Types of Skin Diseases Using
Case-based Reasoning with Image Processing.
International Journal of Conceptions on Computing and
Information Technology Vol. 3, Issue. 3, October’
2015; ISSN: 2345 - 9808.
Batyrshin, I. Constructing Time Series Shape Association
Measures: Minkowski Distance and Data
Standardization. 2013 BRICS Congress on
Computational Intelligence & 11th Brazilian Congress
on Computational Intelligence.
Case-based Reasoning for Skin Diseases Diagnose using Minkowski Distance
201
Bhuiyan MA, Miah MBA, and Mia MR. Detection of
Cancerous and Non-cancerous Skin by Using GLCM
Matrix and Neural Network Classifier. International
Journal of Computer Applications (0975 - 8887).
Volume 132 - No.8, December 2015.
Binbin Lu, Martin Charlton, Chris Brunsdon & Paul Harris.
(2015). The Minkowski approach for choosing the
distance metric in geographically weighted regression.
International Journal of Geographical Information
Science, DOI: 10.1080/13658816.2015.1087001.
Du, WS. Minkowski-type distance measures for
generalized orthopair fuzzy sets. Int J Intell Syst.
2018;1–16. https://doi.org/10.1002/int.21968.
Hay RJ, Johns NE, Williams HC, Bolliger IW, Dellavalle
RP, Margolis DJ, Marks R, Naldi L, Weinstock MA,
Wulf SK, Michaud C, Murray CJL, & Naghavi M.
(2014). The global burden of skin disease in 2010: An
analysis of the prevalence and impact of skin
conditions. Journal of Investigative Dermatology, 134,
1527-1534.
Job CK. (1994). Pathology of leprosy. In: Hasting RC (ed).
Leprosy, 2nd ed., Edinburg, Churchill Livingstone, p
193-224.
John P. Van de Geer. Some Aspects of Minkowski
Distances. Leiden University, Department of Data
Theory, 1995.
José M. Merigó & Montserrat Casanovas. (2011). A New
Minkowski Distance Based on Induced Aggregation
Operators. International Journal of Computational
Intelligence Systems, 4:2, 123-133, DOI:
10.1080/18756891.2011.9727769.
Karakoc E, Cherkasov A, and Sahinalp SC. Distance-based
algorithms for small biomolecule classification and
structural similarity search. Bioinformatics. Vol. 22
No. 14 2006, pages e243 - e251.
Karimkhani C, Dellavalle RP, Coffeng LE, Flohr C, Hay
RJ, Langan SM, Nsoesie EO, Ferrari AJ, Erskine HE,
Silverberg JI, Vos T, & Naghavi M. (2017). Global
Skin Disease Morbidity and Mortality: An Update
From the Global Burden of Disease Study 2013. JAMA
Dermatol, 153, 406-412.
Khadidja B, & Mostefai Sihem. (2015). Similarity
measures for Content-Based Dermoscopic Image
Retrieval: A comparative study. 1-6.
10.1109/NTIC.2015.7368761.
Li B, Chang E, and Wu Y. Discovery of a perceptual
distance function for measuring image similarity.
Multimedia Systems 8: 512 - 522 (2003). DOI:
10.1007/s00530-002-0069-9.
Pal SK, and Shiu SCK. (2004). Foundations of Soft Case-
Based Reasoning. John Wiley & Sons Inc, New Jersey.
Perlibakas, V. Distance Measures for PCA-based face
recognition. Pattern Recognition Letters 25 (2004) 711
– 724.
Singh A, Yadav A, and Rana A. K-means with Three
different Distance Metrics. International Journal of
Computer Applications (0975 - 8887). Vol. 67
Number 10, April 2013.
Slade, Stephen. (1991). Case-Based Reasoning: A Research
Paradigm. AI Magazine. 12. 42-55.
10.1609/aimag.v12i1.883.
Sreedevi E, Padmavathamma MA. Threshold genetic
algorithm for diagnosis of diabetes using Minkowski
distance method. International Journal of Innovative
Research in Science, Engineering and Technology, Vol.
4, Issue 7, July 2015.
Xian S, Sun W, Xu S, and Gao Y. (2014). Fuzzy linguistic
induced OWA Minkowski distance operator and its
application in group decision making. Pattern Anal
Applic, DOI: 10.1007/s10044-014-0397-3.
ICVHE 2019 - The International Conference of Vocational Higher Education (ICVHE) “Empowering Human Capital Towards Sustainable
4.0 Industry”
202