Fuzzy-Based Application Model and Profile Matching for Recommendation Suitability of Type 2 Diabetic

—Diabetes Mellitus (DM) is a metabolic disease characterized by hyperglycemia due to insulin secretion abnormalities and a global health threat. DM has several types, namely type 1, 2, gestational, and other types. Type 2 diabetes patients have the largest number in the world. DM therapy can be done in 2 (two) ways: improving lifestyle and administering drugs. The problems and risks in recommending drugs are essential in the patient's healing process because they are likely to take medicine for life. Approximately 260,000 patients with type 2 diabetes experienced medication errors in 2017. The doctor's mistake in recommending drugs causes a long healing process and costs more. Recommending drugs requires pharmacological knowledge, and not all hospitals have pharmacologists. Several researchers have researched recommendations for antidiabetic drugs, but no studies have yet been found that discuss recommendations for combination antidiabetic drugs for type two to determine dosage and frequency. The number of medications used is 6 to 7, with many parameters 5 to 8. The latest endocrinology guidelines for 2020 state that in recommending antidiabetic drugs, not only 6 to 7 participants, but still need to maintain other aspects. Therefore, this study aims to build an expert system model with a new approach in recommending antidiabetic drugs with more complete parameters and recommend dosage and frequency. The model developed uses the Fuzzy Profile Matching method. Fuzzy is used to calculate the suitability between the patient's condition and the type of antidiabetic drug. Profile Matching is used to calculate the core factor and secondary factor to obtain each drug's total value. The dose was calculated using the FIS Tsukamoto for inputting low dosage, and high dosage calculated the weighted average value. Determination of frequency using the IF-Then function. Model evaluation is done by comparing recommendation data from doctors. The results of the evaluation of the model obtained an accuracy of 90%. This system will reduce medical personnel errors in recommending antidiabetic drugs that can positively impact patients' time, the healing process, and costs. This study provides knowledge that antidiabetes drugs' determination requires many parameters, while other studies used only 4 to 8. This study also provides an overview of the dosages of drugs that drug companies can produce. Usually, the company only makes low and high dosage. This study shows that creating multiple drug dosage is more efficient for patients.


I. INTRODUCTION
Diabetic Mellitus (DM) Type 2 is a group of metabolic diseases with hyperglycemia characteristics that occurs because of an abnormality receptor insulin that lasts long also affects its secrecy. DM type is classified into 4 (four) groups, namely Type 1 DM, type 2 DM, gestational DM, and other type DM [1] [2]. Blood glucose levels are expressed as diabetic, among others, with a rate of HbA1c > 6.5% (mmol/L) [3]. Until today DM is still one of the global health threats. Epidemiological research indicates the tendency to increase the incidence rate and prevalence of type 2 Diabetic Mellitus in various parts of the world [4]. The majority of DM is predicted to grow 3 (three) times in 2030. This increase has been expected by the World Health Organization (WHO) that the year 2030 will reach 21.3 million [1], and Predicted from the International Diabetic Federation (IDF) in 2045 will reach 16.7 million [3].
DM can occur in patients accompanied by other diseases. DM therapy can be done 2 (two) to improve the lifestyle and Drug Administration [2]. Treatment of medications using Oral and Insulin types [5]. Commonly used oral drugs are types of Sulfonylurea, Glinide, Biguanide, Tiazolidin, Alpha Glucose inhibitors, GLP-1, SGLT-2, DPP-4, while for Insulin there are Lispo, Aspart, Glulysine and Faster Aspart [6]. The goal of therapy in DM is to reduce hyperglycemia symptoms, reduce the onset and development of complications, reduce mortality, and improve life quality [6]. Antidiabetic drugs usually pay attention to age, comorbidities, risk of hypoglycemia, and many other factors [7].
Efforts to manage DM still have obstacles in terms of service and health financing [4]. It should be noted that health workers in carrying out their work require high pharmacological accuracy and knowledge [8]. Around 260,000 patients with diabetes experienced medication errors in 2017 [8]. Ignorance and negligence of action to the patient will have an impact on patient safety. One thing that must be considered is the procedure for administering injectable and oral drugs. Giving injection drugs is more at risk of causing hypoglycemic conditions that are dangerous for patients. In addition to economic wastage, irrational drug use patterns can decrease treatment services quality, increase drug side effects, increase treatment failure, and increase insulin resistance [9]. Cases in various health institutions were found to be incorrectly given unnecessary drug combinations. The selection of an appropriate oral hypoglycemic drug is crucial to the success of diabetic therapy, depending on the severity and condition of the patient. Oral hypoglycemic pharmacotherapy can be done using one drug or a combination of two types of drugs [7].
Sub-therapeutic drug administration results in ineffective drug therapy. Drug administration with excessive dosage results in hypoglycemic effects and the possibility of toxicity [10]. Inappropriate use of Insulin often results in hypoglycemia and can lead to weight gain. Unwanted drug effects can occur in long-term use, such as lipodystrophy or loss of fat tissue at the injection site, and allergic reactions can occur, including edema [11]. Treatment must be started as early as possible to prevent or slow the progression of betacell failure in people with impaired glucose tolerance [4].
Several researchers have conducted research that discusses antidiabetic drug recommendations. In the study showed Rung-Ching Chen et al. [12], the drug recommendations used the SWRL technique with 6 (six) types of antidiabetic drugs Metformin, DPP4, Sulfonylurea, Glinide, Thiazolidinedione, Alpha-Glucosidase (AGI) with 6 (six) parameters of HbA1c, Hypoglycemia, Renal, Heart, BMI, and liver. This research was developed with the Fuzzy method that can display the results of drug recommendations based on the most appropriate level of choice [13]. Drug recommendations are also carried out using Fuzzy-TOPSIS with 7 (seven) types of drugs and 8 (eight) parameters [14]. In 2018 Fuzzy, combined with MULTIMOORA with input data scoring, recommended antidiabetic drugs using 8 (eight) parameters. Several researchers have researched recommendations for antidiabetic drugs, but no studies have yet been found that discuss recommendations for combination antidiabetic drugs for type two to determine dosage and frequency. The number of medications used is 6 to 7, with many parameters 5 to 8. The latest endocrinology guidelines for 2020 state that in recommending antidiabetic drugs, not only 6 to 7 participants, but still need to maintain other aspects such as glucagon secretion (Cell Alpha Pancreas), insulin secretion (Cell Beta), glucose fat, glomerular filtration, muscle glycogen and contraindications with pregnant or nursing women and infections [15]. Drug recommendations must be adapted to the patient's condition or variables to avoid errors and drug side effects. The number of patient variables has the main and second variables [16]; therefore, the Profile Matching (PM) method is very appropriate because it has a Core Factor and Secondary Factor calculations.
The problem and the risk of recommending drugs are essential in healing patients to maintain health services quality [10]. This research supports this research; this study aims to build an expert system model with a new approach to recommending antidiabetic drugs with more complete parameters and recommend dosage and frequency. The model developed uses the Fuzzy Profile Matching method. Fuzzy is used to calculate the suitability between the patient's condition and the type of antidiabetic drug. Profile Matching is used to calculate the core factor and secondary factor to obtain each drug's total value. Model evaluation is done by comparing recommendation data from doctors. A safe treatment system needs to be developed and maintained to ensure that patients receive good drug services due to the increasingly varied drugs and the increasing number of drugs and types of antidiabetic drugs [17]. This study's results can be used as an alternative to help paramedics. Young doctors recommend the right dosage and frequency of medicines to improve the quality of health services, accelerate the healing process, and reduce medical costs.

II. MATERIALS AND METHOD
The application of the suitability of antidiabetic drugs to the patient's health condition was developed by illustrating the proposed model's architecture. The development of the model consists of 2 (two) main parts, namely the development knowledge base and development environment presented in Fig. 1 model was developed from the drug suitability model [16].

A. Development Stages
The first development from the expert consultation stage and the result is presented in Figure. 2. Expert consultation was carried out by specialists in internal medicine, diabetes, and pharmacology to obtain parameters and knowledge base. The next step is the process of matching antidiabetic drugs to the patient's condition using a membership curve. The next match's result was calculated by the core and secondary factors using the Profile Matching method. In addition to the type of drug, for determining the dose using Tsukamoto FIS. The stages of development can be seen in Fig. 2.

B. Expert Consultation
Based on consultations with internists and pharmacologists, as well as a review of several works of literature [5], [18], [19], [4], there are 17 (seventeen) parameters that influence determining the delivery of antidiabetic type 2 drugs. In addition to considering the patient's health parameters, the drug's efficacy and price are presented in Table I.

C. Expert System Knowledge Base
The parameters used are made in the form of a knowledge base for each parameter's degree of compatibility with the type of antidiabetic drugs. The knowledge base is presented in Table II. Almost all type 2 diabetic drugs should not be given to DMT2 patients with impaired liver or kidney function, liver, high blood pressure, and severe heart problems. Patients with T2DM aged ≥60 years and overweight (BMI) should be aware of the onset of hypoglycemia. There are types of contraindicated drugs in patients with impaired renal function with LFG ≤ 30 mL/ [4]. Also, drug administration needs to be considered for pregnant or breastfeeding patients and have infections [10].

D. Fuzzy Membership Functions
Based on the knowledge base in table II, they then made in the form of curves and fuzzy logic membership functions for each parameter with the suitability of the type of antidiabetic drug. Curves and membership functions of the kind of antidiabetic drug Biguanide are shown in Table III.  Membership functions need to be made for the types of antidiabetic drugs Sulfonylurea, Glinid, Thiazolidinedione, Alpha-Glucosidase, GLP-1, SGLT-2, DPP4, and Insulin need to be made. Based on the membership function in Table III, the value of each parameter is then calculated.

E. Core Factor and Secondary Factor
Parameter grouping is divided into 2 (two), namely Core Factor (CF) and Secondary Factor (SF). Core Factor is the leading parameter group where the determination of the type of drug given is very dependent on the parameters in this group, whereas a Secondary Factor is a parameter group that does not have a strong influence on the determination of the type of drug given to patients [27]  Calculate the value of CF using a formula:  The grouping core factor's value average value multiplied the weight of 75%, and the secondary factor bore with a weight of 25%. The result of the core factor and secondary factor weights are then added to get a matching value:  This model can evaluate the suitability of the patient's condition with various types of antidiabetic drugs.

F. Dosage and Frequency Drug
The dose and frequency of drug administration are very influential in the therapeutic effect of the drug. Giving excessive dosage, especially for drugs with a narrow range of therapy, will be very at risk of side effects. Conversely, a too small dose will not guarantee the achievement of less than optimal therapeutic levels [17].

G. The domain of Medicine Dosage
Determination of the dose using the parameters in Figure 4 (a). Each parameter becomes an input variable, divided by 2 (two) in linguistic and domain variables. The environment's output is a dose calculated using Tsukamoto's FIS to calculate a more appropriate dosage.  Based on Table IX. The next step is to make a curve for each parameter presented in Fig. 3-5, and the output curves for drug dosages are shown in Fig. 6.  Each parameter's membership value is then calculated based on the membership curve and function, as in Table X.

H. Fuzzy Implication Rules for Dosage
Monotonous fuzzy rules are used as a basis for fuzzy implication techniques. The number of practices used is calculated based on the number of criteria and sub-criteria [28]. The parameters used are 6 (six) as HbA1c, Age, BMI, Renal, Liver, Hypoglycemia, and sub-criteria of each criterion are 2 (two), so the number of rules use is 2 6  Then value z calculation will be performed to look for output using FIS Tsukamoto from each rule given explicitly (crisp) based on α-predicate (fire strength). In this calculation, not all α and z1-64 rules are displayed. The final result is obtained using a weighted average. Examples of the use of practices as follows: = High -(α64 * (High-Low) ) = 3000 -(0.16 * (3000-500)) = 2600

I. Determining Dosage
After a combination of forming rules, the next step is doing a calculation to get the value of defuzzification by adding the rules to regulations 64 to get the weighted average values  Table XI, based on the results of the system recommendations for the correct dosage given by patients as many as 2160 mg/dl.  Fig. 7 The daily dose of medicine recommended by doctor and system Fig. 7 shows the system recommendations can provide daily dosage according to the patient's severity, while the doctor's recommended dosage begins using a low dosage [29]. Giving too low a dosage can result in suboptimal results [17], and recovery is prolonged for up to 1 year. However, for patients receiving the system's recommended daily dose, recovery duration is shorter to ≤3 months [30].

J. Determine Drugs Frequency
The low frequency of use will result in a healing process and have an extended usage interval frequency of drug use that can cause side effects that can worsen the patient's condition. The dose should consider the HbA1c level <8% to determine the drug dosage and frequency [18]; we need proper consideration in determining the dosage and frequency. The frequency of administration of antidiabetic drugs using IF-Then about HbA1c levels shown in Table XIII.

K. Expert System Application
This application uses fuzzy-profile matching, which was built using the Pascal programming language with the Delphi IDE. The application interface can be seen in Fig. 8. Fuzzy logic calculates the value of the match between the patient's condition with the type of drug and profile matching as an inference to display the total amount of each kind of medication. The dose was calculated using the FIS Tsukamoto for inputting low dosage, and high dosage calculated the weighted average value. Determination of frequency using the IF-Then function. Doctors or medics will use this application by inputting several parameters, and the system will display the match values of each antidiabetic drug. Also, the system can communicate as well as the frequency of administration of the appropriate medication

L. Comparison with Existing System
Table XIV shows the differences between several studies of antidiabetic drug recommendations with this study. The difference between this study and previous research is that this study uses more complex parameters to recommend the type of drug and its name. Also, being able to calculate the dosage and frequency based on parameters so that the dose and frequency are more precise and consider the price and efficacy of the drug

A. Recommendation Doctor with System
The data used were 20 test data taken from patients' medical record data at the Bumi Waras Hospital in Bandar Lampung, Lampung, Indonesia, in 2019. Medical record data were calculated using the ordinal scale 1 and 0, as shown in Figure 9. in mapping the suitability of the patient's condition with antidiabetic drugs. The calculation uses a database query by creating a table; then, the selection is based on each patient's condition stored in the view. Data in the next statement is calculated using a query formula to get the total. The results of the query calculation in Figure 10  Calculations using an ordinal scale have weaknesses because they do not produce flexible values to affect the quality of drug recommendations [16]. For example, antidiabetic Sulfonylurea is used for ≤60 years. If calculated using an ordinal scale, patients who are 61 years old cannot be given the type of Sulfonilurena drug, even though up to 65 years of age can still be given the medication. Therefore we need a more flexible calculation using Fuzzy logic [16].
Compared with Ordinal scale calculations, the application of fuzzy logic produces drug recommendations that approach the dataset; this is because fuzzy logic can provide flexible values to provide better anti-diabetic drug recommendations. Based on the number of recommended first-line antidiabetic drugs, Biguanide (Metformin), while for the second-line Insulin. This is according to management guidelines for type 2 Diabetes Mellitus [18].

B. Evaluation of drugs administration
In Antidiabetic drug recommendations, the accuracy of the system is crucial [33]. The course will display all the results, and the doctor will choose the best based on expertise. Evaluate the suitability of drugs recommendations based on the system, and the doctor, True Positive (TP) is used, which means the doctor approves the recommended drug. The dataset (DS) is the total amount of data, the formula shown in Table XVII. The first stage of testing compares drug recommendations using the Ordinal scale, and the second stage will be carried out to compare drug recommendations using fuzzy logic. The results of drug recommendations using the Ordinal scale can be seen in Table XVI.  The recommendation to use Fuzzy does not have much difference with the dataset doctor. The difference lies in the number of Biguanide recommendations that the dataset recommends as many as 14, but the system only recommends 12. Based on the accuracy value calculation, the fuzzy logic application has better accuracy, with an average difference of 43%. The application of fuzzy logic was high-speed and lower cost in recommending reliable drugs [26].

IV. CONCLUSION
Based on the description, explanation, and testing that have been done, we get a few conclusions. This study applied antidiabetic drugs' suitability based on the patient's health condition using the Profile Matching and Fuzzy Logic methods. Based on the evaluations Fuzzy Logic can recommend antidiabetic drugs that are better than using the Ordinal scale. In addition to the recommendation of the type of medicine, the system can also recommend the dosage and frequency of using Tsukamoto's FIS so that it is more precise and reduces the errors of medical staff in recommending drugs and can have a positive impact on patients in terms of time, the healing process, and lower costs. This study provides knowledge that antidiabetic drug determination requires as many as 17 parameters, while other courses only use 4-8 parameters. This study also describes the number of drugs that drug companies can produce. Usually, companies only make low and high dosage. This research shows that creating various dosages of the drug is more efficient for patients. However, this research still needs to be reviewed and continued considering that it still has some weaknesses and shortcomings from the dataset to the number of parameters.