You are on page 1of 7

MITS5509

Intelligent Systems for Analytics Assignment2

Data Mining Based Hybrid Intelligent System for Medical Application

By

Adane Nega and Alemu Kumlachew


Research Report

Student Name

Student ID
Table of Contents
1. Introduction.........................................................................................................................................1

2. Key Points Mentioned in Article on Intelligence Systems for Analytics.............................................1

3. Conclusion...........................................................................................................................................4

Reference.....................................................................................................................................................5
1. Introduction
Computer applications have a considerable impact on the healthcare industry. Artificial
Intelligence (AI) is one of the technologies used in healthcare applications including monitoring
of patients, administration of drug dosage, modeling the functions of the brain, speech analysis,
and administration of drug dosage. It helped in developing a new system called Hybrid
Intelligent System (HIS) to solve the problems in medical treatment for various diseases. The
major purpose of the present report is to perform a review of the article related to the use of data
mining techniques in Intelligence Systems. The article “Data Mining Based Hybrid Intelligent
System for Medical Application” [ CITATION Ada17 \l 1033 ] is considered for reviewing. It helps
to improve knowledge on applying data mining techniques in intelligence systems. This report
will cover the key findings produced in the article including the knowledge acquisition for HIS
including traditional and data mining techniques, HIS architecture, evaluating and testing the
cases, and integrating the knowledge discovery rules into HIS.

2. Key Points Mentioned in Article on Intelligence Systems for Analytics


In the article, the authors focused on the various concepts related to using data mining in HIS
that are used in medical applications to provide solutions to diagnosis problems.

AI uses two types of techniques to solve medical problems including Case-Based Reasoning
(CBR) and Rule-Based Reasoning (RBR). CBR and RBR both techniques are useful in analyzing
the patient health conditions and providing suitable recommendations to treatment by dealing the
industry knowledge and medical information. These are further helpful in reducing medical
errors and complications, enhancing the compliance of patients, and increasing the quality of the
decision-making process. The authors focused on the application of data mining algorithms to
the considered disease called Tuberculosis (TB) using the HIS. Data mining is defined as the
extraction of useful information collected from different sources by applying various algorithms.
Its applications are identified in various areas including the financial industry, biological data
analysis, and forecasting of weather conditions.

HIS is the association of data mining and AI used in healthcare. It is used in resolving complex
business problems in the medical field. It is also used in the classification of medical data and
solving health problems including diabetes, live disorder, and breast cancer. To analyze the

Student ID Student Name 1


performance of this system different metrics such as sensitivity, correctness, and specificity in
healthcare applications[ CITATION Man14 \l 1033 ] . Medical practitioners are accepting this system
as an effective supporting tool in making major decisions. Knowledge acquisition is the key
problem presented in HIS due to scarcity of knowledge among the knowledge engineer and
domain expert on techniques and inability to correct and describe the knowledge. The AI used in
the hybrid intelligent system has several benefits and drawbacks. The HIS system is influenced
by the performance and cost of the AI and the type of methods used for gaining knowledge.

Knowledge acquisition is an important aspect of developing the HIS. Two types of methods
available for knowledge acquisition including the traditional approaches and data mining. The
data mining technique supports in overcoming the challenge of knowledge acquisition.
Traditional methods applied to analyze the patient conditions for a Case-based approach. Data
mining techniques and classification algorithms applied for Rule-based approach. But, most of
the users adopting the rule-based approach in the process of knowledge acquisition to make the
system effective. It is effectively using in the prediction of the chances of heart diseases using
different data mining techniques including the neural networks, naves Bayes, and decision
trees[ CITATION Ami18 \l 1033 ].

Traditional methods used are identified as document analysis on the past cases related to diseases
and conducting the interviews with the professionals associated with TB disease treatment. Both
methods are used in primary diagnosis and treatment accessing the domain and general
knowledge. Data mining is the extraction of data unknown and interesting patterns from data. It
resolves problems of the discovery of knowledge and prediction. Data mining techniques have
some advantages over traditional approaches. Those are removing the bottleneck of knowledge
acquisition, reducing the errors, and generates high quality of knowledge in fewer costs. Even
though, data mining is automated requires the support of domain expert. The medical profile of
patients is used as a basis including age, blood sugar, blood pressure, and gender. HIS is also
extended to dangerous diseases like cancer.

The process of data mining with AI follows various steps to produce knowledge from a large
volume of data sets. Those include selecting the data type required including the weight, chest
pain, fever, age, weight loss, lack of appetite, type of TB, and different type of tests, applying
cleansing techniques to correct and detect the error data, integrating the data from a different

Student ID Student Name 2


type of sources, selection of important attributes for performing data analysis, and selection of
data mining algorithm and model. The authors conducted four experiments applying the
classifier and decision tree including BFTree, J48, PART, and JRIP. BFTree and J48 algorithm
have belonged to tree-based classifiers and PART and JRIP have belonged to rule-based
classifiers. Production rules are established from the knowledge created to understand the
diseases. The performance of these techniques is evaluated effectively and identified that the
PART classifier showing high performance in creating the knowledge base to automate the
process of knowledge acquisition.

HIS depends on combined RBR and CBR to generate knowledge. Its architecture is comprised of
a set of modules including integration module, decision module, user interface, and data mining.
In a data mining module, different activities are performed including the data selection, pre-
processing of data, valuation of the model, and applying the rules. The integration module uses
the RBR to represent the knowledge and inference engine and an explanation facility is used to
build the knowledge base. The decision module depends on CBR to solve the problem of
diseases by testing the cases and to provide a solution and it associates the CBR and RBR
modules. The CBR module performs the function of gathering and choosing the cases of
patients, developing, demonstrating, and recalling the case in the database. If it doesn’t give a
solution, the problem is transferred to RBR. The RBR considers rules and facts on the basis of
knowledge to facilitate recommendations to treatment. Then, the solution is displayed on the user
interface to provide it to users[ CITATION Ren16 \l 1033 ]. The user interface offers various benefits
such as displaying the laboratory results and disease symptoms. Based on the past cases of
patients, effective solutions are provided by the domain experts to physicians.

Testing and evaluation is a key problem for Hybrid intelligent system. Two types of tests are
used to generate an accurate and acceptable solution for the treatment of TB including the
performance testing and user acceptance testing. The main reason for conducting user acceptance
testing is determining the satisfaction levels of end-users. System performance testing is helpful
in evaluating the opinions of experts and providing an accurate response. Testing is performed
on RBR and CBR modules as these modules are executed individually and the testing procedures
are different for these two modules [ CITATION Isa06 \l 1033 ]. Testing on RBR is conducted with
six experts in the healthcare domain. The test cases are equally distributed to them to find out the

Student ID Student Name 3


incorrectly and accurately diagnosed disease cases. Precision and recall are used in testing the
CBR module.

The accuracy of diagnosis is determined by the following formula

Total test casesdiagnosized accurately


Accuracy of diagnosis= X 100
∑ of total test cases
For this HIS system, the accuracy rate is determined at 89.4%. This indicates that the system is
effective in generating knowledge applying the data mining techniques to provide treatment and
diagnosis for TB. The average performance of the system is obtained as 87.5% and this indicated
that data mining is an efficient technique in treating TB.

HIS supports a lot in curing the TB disease by applying the data mining techniques and it is able
to provide solutions to critical medical issues based on existing knowledge. The article
concluded that the bottleneck presented in knowledge gaining is removed completely and
improvements required to HIS including improving the user interface to provide language
preferences and updating the rules presented in the knowledge database. It is identified that the
data mining technique is the effective feature of the HIS in building high-quality knowledge at
lower costs.

3. Conclusion
In the present report, the review is effectively performed on the article to understand how data
mining is used in HIS. It included various key points discussed in the article by the authors.
Those include use of AI in medical areas, understanding the functions used for knowledge
acquisition for HIS, discussion of traditional and data mining techniques by comparing with each
other, the process of data mining with clear steps, architecture of HIS system including different
modules to build the knowledge, analyzing the two types of tests conducted on RBR and CBR
modules and how these will generate the solutions to medical problems. From this, it is
understood that data mining has lot of significance in intelligence systems.

Student ID Student Name 4


Reference

Student ID Student Name 5

You might also like