Implementation of Decision Tree and Dempster Shafer on Expert System for Lung Disease Diagnosis

The expert system is a computer system that contains set of rules to solve problems like an expert. The lungs are one of the vulnerable respiratory organs. The purpose of this research is to implement decision tree and dempster shafer method on lung disease diagnosis and measure the accuracy of the system. The symptom was searched using forward chaining decision tree and the diagnosis was calculated using dempster shafer method. Dempster Shafer method calculates the possibility of a lung disease based on the density of probability value that possessed by each symptom. This research used 65 data obtained from medical record of Puskesmas Tegowanu Grobogan Regency. General symptoms and types of disease are used as a variable. Based on the results of the study, it can be concluded that the results of the diagnosis using dempster shafer method has an 83.08% accuracy.


INTRODUCTION
Expert systems are part of the high-level software or high-level programming language which attempt to duplicate the expert functionality in a particular area of expertise [1].It can be used to overcome multiple problems by giving some advice like an expert knowledge [2].In the application, the problem that needs to be solved is not only the algorithm but also the expert-field problem that is difficult to understand [3].There are a knowledge base and an inference system (set of rules) [4].In other words, the expert system is software-based systems that create or evaluate decisions based on rules defined in the software [5].The purpose of the expert system is not to replace human role, but to represent a human knowledge into a system form, so that it can be used by many people [6].From some previous studies, expert system provides good results for solving cases that use complex data, such as skin disease diagnosis, pregnancy illness diagnosis, asset damage analysis and digestive disease diagnosis [7][8][9][10].
The Dempster Shafer theory was introduced by Dempster (1967) and expanded by Shafer (1976).Dempster Shafer is a representation, combination, and propagation of uncertainty that has some institutive characteristic according to the way the expert thinks, but it has a strong mathematical basis [11].This theory can be interpreted as a general form of probability theory, which is used for the dataset, not for single data [12].In addition, this theory aims to represent and overcome the uncertainty of information.The most important about dempster shafer is the ability to combine different data sources to improve the quality of information [13].
Someone who is suffering from a disease with cough symptoms would need to consult with a doctor so that he immediately know the illness that he suffered [14].Lung disease is a disease associated with the respiratory system in humans, can be bad if not treated immediately seriously.The lungs served as a place of exchange of oxygen needed by humans and release carbon dioxide which is the result of the rest of the respiratory process that must be removed from the body so that the body's need for oxygen remains fulfilled [15].A doctor can diagnose the disease by analyzing the patient symptoms.But, the presence of technological advances makes it can be diagnosed more quickly by using an expert system [16].
Based on the description, the purpose of this research is to implement decision tree and dempster shafer method in Expert System Diagnosis of Lung Disease and measure the accuracy of that.

Decision Tree
The decision tree is a classification method that uses a representation of a tree structure, each node represents an attribute, a branch represents the value of an attribute, and the leaf represents the class.The top node of the decision tree is referred as a root [17].The decision tree is decision support tools that use hierarchical tree structures to classify the classes based on a set of questions [8].The decision tree consists 3 types of nodes: 1. Decision Node: generally represented by a box.2. Node of opportunity: generally represented by a circle.3. End node: generally represented by a triangle.
Decision tree as a decision support tools can provide effective decisions because they have several advantages such as: a) Easy to understand and interpret.b) Has a value even with small amount of data.c) Can be combined with other decision-making techniques.d) Exposing all problems so that all possibilities can be classified.e) Allows to analyze in making decisions about the possibilities of some alternatives f) Provide a framework for measuring the result of value and probability of reaching a decision.g) Help to make the best decisions based on available information.

Dempster Shafer
Dempster shafer is a mathematical theory for looking evidence based on belief function and plausible reasoning, used to combine separate information (evidence) and calculate the probability of an event.In other words, Dempster shafer is a mathematical theory for facts [12].There are various kinds of reasoning with complete and consistent models, but in the reality, there are many problems that can't be resolved completely and consistently.The inconsistency is due to the addition of new facts.This is called as nonmonotonic reasoning.Dempster shafer is able to overcome this inconsistency [18].In the expert system, a disease has a number of evidences that will be used in the uncertainty factor in making a decision to diagnose a disease.To overcome a number of evidences, dempster shafer using a rule that known as Dempster's rule of combination.

Process of Planning
The steps of developing an expert system diagnosis of lung disease are shown in Figure 1.

System of Design
The design of this expert system using waterfall model.A waterfall is an approach based on the assumption that major decisions must be made before encoding begins [19].This model is often used by systems analyst [20].There are four stages in the waterfall model [21].1) requirement analysis stage is defining the entire software format, identifying all the needs and outlines of the system [22].
2) The design stage is designing application includes the interface design and database structure design [23].To create an interesting web-based application program (website) it must be designed beforehand, so the achieved results are suitable with the predetermined objectives [24].
3) The implementation stage is designing software which realized as a series of program or program unit [25].4) The testing stage is to test whether the system is ready and feasible to use.The (1) tester can define the set of input conditions and perform testing on functional specifications of the program [26].

Data Collecting
The data used in this study is the result of medical records of 65 patients in 2016 who suffered from lung disease including Tuberculosis, Asthma, Bronchitis, and Pneumonia.This data are taken from Puskesmas Tegowanu Grobogan.This data contains the symptoms and doctor diagnosis results of patients in Puskesmas Tegowanu.

Interview
The interview was conducted with a Tegowanu Public Health Center internal medicine specialist.This interview gives the researcher a set of weight value from each symptom in lung disease shown in Table 1.

Testing
After testing 65 medical record data of lung disease patients at Puskesmas Tegowanu Grobogan, the result obtained accuracy of 83,08% as seen in Table 2.The testing of medical record data using a system provides 54 accurate data out of 65 data.The rest 11 data is less accurate due to the system with the dempster shafer method uses a strong calculation, so it doesn't easily influenced by other factors.While the specialist in internal medicine as an expert has other considerations in determining the disease suffered by the patient.

Implementation
A consultation page is a page that users can use to consult.The user can consult with the system without login.In conducting the consultation, the user fills in the name and any symptoms experienced.Figure 2 shows the consultation page.The results page of the consultation is a page that provides diagnostic information from the input symptoms.This page shows the patient symptoms experienced, the treatment advice and his diagnosis.Figure 3 shows the results page of the consultation.

CONCLUSION
Dempster shafer method and decision tree is a method to calculate the uncertainty of a problem, this uncertainty is due to the addition of new facts.Dempster shafer can optimize the diagnosis that is produced because the system is not only based on the rule but has a value.The level of trust is more accurately supported by the decision tree as a supporter in disease prediction.Dempster shafer works by using the density value or weight value of each known fact, derived from an expert.The weights of these factare combined to produce a combination of known factual density values.The end result is a fact or a combination of facts with the greatest weight value.After

Figure 1 .
Figure 1.The Steps of Developing An Expert System

Figure 3 .
Figure 3. Admin and Expert Diagnosis Result Page