COMPUTER SCIENCE
Abbas F. H. Alharan,
Computer Department, College of Education for Girls, University of Kufa, Kufa, P.O. Box (21),
Najaf Governorate, Iraq Nabeel Salih Ali
Information Technology Research and Development Centre, University of Kufa, Kufa, P.O. Box (21),
Najaf Governorate, Iraq Zahraa M. Algelal
Computer Department, College of Education for Girls, University of Kufa, Kufa, P.O. Box (21),
Najaf Governorate, Iraq DOI: 10.24412/2520-6990-2022-19142-4-10 AN ADAPTIVE NEURO-FUZZY INFERENCE SYSTEM AND PRINCIPAL COMPONENT ANALYSIS: A HYBRIDIZED METHOD FOR VIRAL HEPATITIS DIAGNOSIS SYSTEM
Abstract
Viral Hepatitis is considered a significant public health challenge and an issue that affects the deaths of approximately 1.34 million people per year and takes man's life. Thus, robust data mining techniques require the diagnosis of the disease by offering a successful interpretation of the medical datasets and are still a significant problem that needs to be addressed by researchers. Different learning methods have been developed for disease diagnosis purposes. Hence, many previous methods in the literature lack the ensemble learning of the clinical data to improve the classification accuracy by hybridizing the outputs for various prediction systems. However, this study proposed a hybridizes the machine learning method based on Principal Component Analysis (PCA) and Adaptive Neuro-Fuzzy Inference System (ANFIS) to Improve the Accuracy of the hepatitis diagnosis. PCA is used to reduce the hepatitis disease features space dimensionality of the selected dataset that is taken from the UCI Repository Databases. Therefore, 19 feature attributes are reduced to 8 using the PCA technique. The reduced features are given to inputs ANFIS classifier to evaluate the accuracy performance. The hybrid PCA-ANFIS method is achieved remarkable higher Accuracy (97.41%) compared with the state-of-the-art, and it can be an adopted tool for diagnosing Hepatitis.
Keywords: Adaptive Neuro-Fuzzy Inference System (ANFIS), Principal Component Analysis (PCA), Attribute Selection, Classification, Hepatitis.
Introduction. Hepatitis is inflammation of hepatocytes in the liver due to infectious viruses and noninfectious agents. Therefore, viral Hepatitis is a most important public health challenge requiring an urgent response, reasoning that approximately 1.34 million die annually [1]. It has five main types of Hepatitis: A, B, C, D, and E. In actuality, B and C lead to chronic disease in millions. They are the main factor in fatalities from viral Hepatitis, liver cancer, and liver cirrhosis. Hepatitis B and C infections are estimated to affect 354 million people, yet for most of them, diagnosis and treatment are still out of reach [2]. Thus, early-stage hepatitis prediction is a significant factor in efficient treatment for disease decisions [3]. Many supervised and unsupervised Machine-learning (ML) techniques have proposed an automatic prediction system in several medical applications [4]. Hence, singly and hybrid learning algorithms are adopted in the literature and achieved significant results for different diagnosis problems [5], [6], [7]. Moreover, attaining a better performance average for disease diagnosis in the medical field remains a critical issue in most research recently [8]. Consequently, prominent researchers have recently used neural networks and fuzzy inference methods in medical diagnostic [9].
In the current study, a hybrid method (PCA-ANFIS) for the diagnosis of hepatitis disease is proposed. The technique combines an adaptive neuro-
fuzzy inference system and PCA to enhance the Accuracy of the hepatitis diagnosis. PCA is used to reduce the feature space dimensionality of the selected UCI dataset and retain the essential information from the data. However, to categorize the hepatitis dataset by the classes, the chosen attributes are sent to the ANFIS. The UCI dataset was gained from the UCI Repository in ML Databases. In the classification phase, these reduced features by PCA will be given to inputs ANFIS classifier to evaluate the accuracy performance. The remainder of the essay is structured as follows: The related studies on hepatitis illness diagnostics are reviewed and criticized in Section 2. While Section 3, the proposed method and materials are presented. Besides, the experimental results are shown in Section 4. Finally, this study's conclusion and future works are stated in Section 5.
Related Works: Exploring new methods by researchers to develop an effective and robust disease detection in human diseases faced challenges and issues. However, a large number of modern techniques in learning methods are presented in the literature. These techniques include single, hybrid, and combined approaches to diagnosing human diseases in the medical field.
According to [18], Nivaan et al. (2020) proposed a logistic regression model on the hepatitis disease da-
taset from UCI-ML. The paper conducted a logistic regression that provides analysis results with a classification accuracy of (83.33%). Besides, Basarslan et al.
(2019) developed a system to aid the decision parameters for clinical persons in hepatitis diagnosis based on correlation and fuzzy-based-rough attribute selection approach. Then data were classified using techniques such as Random-forest, K nearest neighbor, Naive-Bayes, and Logistic regression. The Accuracy obtained using the Random Forest algorithm (84.9%) is better than other methods [17]. Likewise, in [16], Sahebi et al.
(2020) suggest providing a feature selection method for dimension reduction, referred to as GeFeS, which is based on a parallel genetic algorithm (GA). This article introduces an improvement of the mutation and crossover operators using a new operator. Different UCI datasets were applied for experimental, and their performance accuracy was (94.28%) with features attributes reduced. The proposed model validated by adopted cross-validation is integrated into the GA process. Besides, the hepatitis disease diagnosis method is presented by Ahmad et al. [14]. The method uses information gain to features reduction and ANFIS to receive the selected attributes to make a hepatitis illness diagnosis. The experimental results achieved (95.24%) accuracy. Likewise, a detecting hepatitis disease method by Sartakhti et al.(2012) in [15] is presented by using SVM and simulated an annealing algorithm (SVM-SA). The (SVM-SA) hybrid method implemented on
UCI clinical dataset with tenfold cross-validation and tuned parameters enhances classification accuracy. The accuracy result of the SVM-SA method is (96.25%). Finally, a work by Elaboudi and Benhlima [19] introduced a model based on the stochastic optimization method and PCA technique. The optimization method includes SVM with CEO (Cross-Entropy Optimization). The classification accuracy of the model was achieved (97.2%).
Materials and Methods: This Section describes the materials and phases for the proposed hybridized method, including details for the PCA-ANFIS system, data preparation and preprocessing principal component analysis (PCA), and Adaptive Neuro-Fuzzy Inference System (ANFIS).
A. Proposed (PCA-ANFIS) Model: In this Section, different steps are presented and described in detail to provide a big picture for the proposing PCA-ANFIS model. The first step in the hybridized method is the preparation and preprocessing of the dataset selected. In this step, the Hepatitis dataset is taken from the UCI machine learning repository and then prepro-cessed. At the same time, the second step aims to extract the valuable attributes from data obtained by the first step using the PCA technique. Lastly, the third step intends to evaluate the classification's performance using the Adaptive Neuro-Fuzzy Inference System (ANFIS). Figure 1 shows the proposed system's flowchart, as shown in Figure 1.
Hepatitis Dataset (Preparation & Preprocessing)
V
Figure 1 Flowchart of the proposed system
B. Data Preparation: The UCI - dataset selected is taken out from the machine-learning repository. The UCI- dataset was collected from Carnegie-Mellon University by Gail Gong (http://archive.ics.uci.edu/ml/da-tasets/Hepatitis). The dataset consists of 155 records and 20 attributes (one of them for the class). Besides,
there are two classes, 123 records (79.35%) for the "live" class and 32 records (20.65%) for the "die" class. Table 1 shows the attributes and their values as in the repository. Table 1 lists all attributes of the Hepatitis UCI dataset.
Table 1
Features "Attributes" of the Hepatitis UCI- Dataset_
No. Name of Attribute Value
1 Class DIE, LIVE
2 Age 10, 20, 30, 40, 50, 60, 70, 80
3 Sex Male, Female
4 Steroid Yes, No
5 Antivirals Yes, No
6 Fatigue Yes, No
7 Malaise Yes, No
8 Anorexia Yes, No
9 Liver big Yes, No
10 Liver firm Yes, No
11 Spleen_palable Yes, No
12 Spiders Yes, No
13 Ascites Yes, No
14 Varices Yes, No
15 Bilirubin 0.39, 0.80, 1.20, 2.00, 3.00, 4.00
16 ALK_phosphate 33, 80, 120, 160, 200, 250
17 Sgot 13, 100, 200, 300, 400, 500
18 Albumin 2.1, 3.0, 3.8, 4.5, 5.0, 6.0
19 Protime 10, 20, 30, 40, 50, 60, 70, 80, 90
20 histology Yes, No
C. Principal Component Analysis: In the classification phase, attribute selection is a critical strategy for determining the importance of attributes connected to the class label [20]. Principal Component Analysis (PCA) is an unsupervised attribute selection technique; it works by projecting a subspace in which the variance of projected data is maximized, or the mean square error between the feature vector and its projection is minimized [21] [22]. In other words, PCA can be defined as determining which features in a set of linearly dependent features are relevant. The "best" eigenvectors of the data covariance matrix can be used to define the "best" low-dimensional space. Principal components are the eigenvectors that correspond to the largest eigenvalues. Suppose xi, x2, x3...xm are N*1 vectors Step1: For each feature, calculate the mean using the following equation:
* = 1^=1*1 (1) Step2: Subtract the mean from each value in the vector:
0i=x-xi (2)
Step3: From the matrix, A=[0i 02...0m ](N*M) calculate:
C=1lX=10n0'
-AAT
Where C is the sample covariance matrix (N*N) which describes the scatter of data.
Step4: Calculate the eigenvalues of C
(3)
C: Xi>X2>"-^N (4)
Step5: Calculate the eigenvectors
C: ui, ui,...,UN (5)
The vector x, which is, (x - x) can be written as a linear combination of the eigenvectors, as the following:
x - x = b1u1 + b2u2 + —+ bNuN (6) x-x = biUi where bt = (7)
'~'l = 1 1 1 1 (Ul-Ul)
Step6: Reduce the features by keeping the ones (K) corresponding to the highest eigenvalues: x-x = Ya=1 biUi where K <N (8) To perform the feature reduction, the linear transformation RN^RK is used as the following:
(x-x) = UT(x-x) (9)
D. Adaptive Neuro-Fuzzy Inference System (ANFIS): An ANFIS is an Artificial intelligence (AI) kind that combines an Artificial Neural Network (ANN) algorithm with a Takagi Sugeno fuzzy logic controller [23]. The ANFIS was firstly introduced by Jang in 1993 [24], and the architecture of ANFIS is explained in Figure 2.
\b1] b2 y-l
-bK- .uTK.
Figure 2 The Adaptive Neuro-Fuzzy Inference System Architecture
«coyyomum-jmtmal» #mre), 2022 / computer science
7
Assume the ANFIS has two inputs (x and y). The first-order fuzzy system consists of two rules (R1 and R2).
Rl: If x is Al and y is Bl, then fl =
= alx + bly + cl (10)
R2:If x is A2 and y is B2, then f2 =
= a2x + b2y + c2 (11)
In R1 and R2, the given inputs are labeled with x, y, and fuzzy sets are with A and Bi; the output is determined by fuzzy rules labeled with f1 and f2, and the linear parameters (consequent part) are indicated by a1, b1, c1, a2, b2, c2.
The structure of ANFIS comprises five layers, as described in Figure 2. The number of neurons per layer is different. The first layer is in charge of decomposing the inputs using a specific number of membership functions (MFs) for each input. Let the output of the A node is indicated by of, of is calculated as the following:
ol = Exp[-(^)2] (12)
Where x is the input, mi and ni are the MFs parameters.
The second layer produces the weight by multiplying the outputs of the first layer. Each node output in this layer denotes the firing strength of a rule (of).
of =Wi= Mx) x nBfy), i = l,2 (13) In the third layer, each ith node computes the ratio with the ith rule's firing strength to the summation of all
rule's firing strength (normalization), as in the following equation.
= .....) (14)
Where of and wL are the outputs from layer 3, named normalized firing strengths, Wi is the output of layer 2.
Every single node (i) in layer 4 is a square of node function, as in the following equation.
04 = wJi = Viz, (ptx + qty + rf)
(15)
Where of and \v'lfi are the output of layer 4, f is the linear functions, Wt is the outputs of layer 3. pt, qh rt are the consequent parameters.
Furthermore, finally, the layer 5 computes the summation of all incoming signals to produce the final output (last single node) as the following:
of = final output = £wifi= (16)
Liwi
Where of and £ vTlfi are the output of layer 5. Experimental Results for PCA-ANFIS
Method: After downloading the hepatitis dataset from UCI, the data were preprocessed for better analysis. The mean value of that attribute replaced the missing data in a specific attribute. Then, the PCA is used to obtain the best 8 attributes from 19 attributes of the pre-processed data. WEKA tool has been used to implement the attribute selection using Ranker search. Table 2 shows the selected attributes with the best eigenvalue (4.446) and ranking (0.766). Table 2 lists the selected attributes of Hepatitis by PCA technique.
Table 2
Selected attributes by PCA
No. Attribute
1 spiders
2 albumin
3 fatigue
4 malaise
5 ascites
6 varices
7 bilirubin
8 anorexia
The data with the selected eight attributes mentioned in Table 2 are recommended as input for ANFIS classification. The data with selected attributes were divided into training (75%, 116 records) and testing (25%, 39 records). Next part, Sugeno fuzzy inference system is applied to construct a fuzzy inference system (FIS). From the input 8 attributes, the FIS generates
member functions (MFs), and MFs generate rules, and the rules generate rules output, rule output generates MF output, and MF output generates the output, which is an individual output value as we can see in Figure 3. The significant advantage of ANFIS is that it benefits of Takagi-Sugeno fuzzy inference system and neural learning capabilities[25].
Figure 3 Constructing of a fuzzy inference system (FIS)
Eight inputs (a1, a2, a3, a4, a5, a6, a7, a8) were passed to the ANFIS that produced the output (z). The fuzzy if-then rule of ANFIS is expressed as the following:
If a1A1 and a2B1 and a3C1 and a4D1 and a5E1 and a6F1 and a7G1 and a8H1, then
fl = kal + la2 + ma3 + na4 + oa5 + pa6 + qa7 + ra8 + t (17)
Where k, l, m, n, o, p, q, r, t are the linear output parameters.
Figure 4 shows the MATLAB ANFIS structure. Thus, to evaluate the (PCA-ANFIS) method, the accuracy measure was calculated as (97.41%). Table 3
According to the performance accuracy of classification results based on the features. The current hybrid method achieved (97.41%) accuracy, which is higher than previous works such as RS-BPNN (97.3%), LFDA-SVM (96.77%), and PCA-LSSVM (96.12%) that belong to articles [28], [29], and [26] sequentially. Moreover, compared with others, it depends on the classification algorithm ANFIS and its accuracy results. The proposed hybrid approach (PCA-ANFIS) also obtained remarkable significant high Accuracy better than (LDA-ANFIS) method in [13] and the (information gain-ANFIS) method in [14] that were its Accuracy (94.16%) and (95.24%) consecutively. Thus, it can be an adopted tool for diagnosing Hepatitis.
Conclusions: Classification accuracy plays a vital role in developing disease diagnosis methods and has been a significant concern for researchers in the literature for such purposes. Several single learning techniques are presented in the state-of-the-art for disease classification. This article aims to diagnose Hepatitis by proposing a hybrid medical diagnosis method combining PCA and the ANFIS to predict Hepatitis disease. The hybrid PCA-ANFIS method used several parameters such as: "attribute", "spiders", "albumin", "fatigue", "malaise", "ascites", "varices", "bilirubin", and "anorexia". Attributes were reduced using the unsupervised PCA technique to gain a better number of features and fed the set of attributes selected to the ANFIS for diagnosing Hepatitis disease.
The discussed method is evaluated using a real-world dataset from the UCI Repository of Machine
shows the comparison between the PCA-ANFIS and other prior works in Accuracy.
Table 3
Learning Databases. The evaluation results of the performance accuracy have shown that the (PCA-ANFIS) method achieved (97.41%) is compared with works that have been done before (see Table 3), in particular, ANFIS classifiers, namely information gain-ANFIS [14] and LDA-ANFIS [13], which obtained (95.24%) and (94.16%) also remarkable accuracy such RS-BPNN [28] for (97.3%) performance accuracy.
Future directions intend to focus on the implementation time reduction and computation by developing the hybrid PCA-ANFIS method. As well as increase the classification accuracy and apply more clinical datasets in the experimental results.
References
1. WHO, "Executive summary - Global hepatitis report, 2017," World Heal. Organ., 2017, [Online]. Available: https://apps.who.int/iris/rest/bit-streams/1082595/retrieve.
2. World Health Organization (WHO), "No Title." https://www.who.int/health-topics/hepati-tis#tab=tab_1.
3. Ali, N. S., Alyasseri, Z. A. A., & Abdul-mohson, A. (2018). Real-time heart pulse monitoring technique using wireless sensor network and mobile application. International Journal of Electrical and Computer Engineering, 8(6), 5118.
4. Abdulkareem, K. H., Mohammed, M. A., Gunasekaran, S. S., Al-Mhiqani, M. N., Mutlag, A. A., Mostafa, S. A.,. .. & Ibrahim, D. A. (2019). A review of fog computing and machine learning: concepts, applications, challenges, and open issues. IEEE Access, 7, 153123-153140.
Figure 4 Structure of ANFIS
Comparison between the PCA-ANFIS and other Prior Works in terms of Accuracy
Ref Method Accuracy (%)
[9] (2006) PCA-AIRS 94.12
[13] (2009) LDA-ANFIS 94.16
[14] (2019) information gain-ANFIS 95.24
[16] (2020) GeFeS 94.28
[26] (2011) PCA-LSSVM 96.12
[27] (2012) CBR-PSO 94.58
[28] (2015) RS-BPNN 97.3
[29] (2011) LFDA-SVM 96.77
[30] (2021) PCA-SVM 93.55
The Proposed Method PCA-ANFIS 97.41
«coyyomum-jmtmal» #mre), 2022 / computer science
5. N. Tanwar and K. F. Rahman, "Machine learning in liver disease diagnosis: Current progress and future opportunities," IOP Conf. Ser. Mater. Sci. Eng., vol. 1022, no. 1, 2021, doi: 10.1088/1757-899X/1022/1/012029.
6. N. C. Caballé, J. L. Castillo-Sequera, J. A. Gómez-Pulido, J. M. Gómez-Pulido, and M. L. Polo-Luque, "Machine learning applied to diagnosis of human diseases: A systematic review," Appl. Sci., vol. 10, no. 15, pp. 1-27, 2020, doi: 10.3390/app10155135.
7. Ali, N. S., & Alyasseri, Z. A. A. (2017). Wireless Sensor Network and Web Application Hybrid Scheme for Healthcare Monitoring. Journal of Soft Computing and Decision Support Systems, 4(5), 1-7.
8. Alharan, A. F., Algelal, Z. M., Ali, N. S., & Al-Garaawi, N. (2021, September). Improving Classification Performance for Diabetes with Linear Discriminant Analysis and Genetic Algorithm. In 2021 Palestinian International Conference on Information and Communication Technology (PICICT) (pp. 38-44). IEEE.
9. K. Polat and S. Güne§, "A hybrid medical decision making system based on principles component analysis, k-NN based weighted preprocessing and adaptive neuro-fuzzy inference system," Digit. Signal Process. A Rev. J., vol. 16, no. 6, pp. 913-921, 2006, doi: 10.1016/j.dsp.2006.05.001.
10. M. Nilashi, H. Ahmadi, L. Shahmoradi, O. Ibrahim, and E. Akbari, "A predictive method for hepatitis disease diagnosis using ensembles of neuro-fuzzy technique," J. Infect. Public Health, vol. 12, no. 1, pp. 13-20, 2019, doi: 10.1016/j.jiph.2018.09.009.
11. V. R. Elgin Christo, H. Khanna Nehemiah, B. Minu, and A. Kannan, "Correlation-based ensemble feature selection using bioinspired algorithms and classification using backpropagation neural network," Comput. Math. Methods Med., vol. 2019, 2019, doi: 10.1155/2019/7398307.
12. M. S. Bascil and F. Temurtas, "A study on hepatitis disease diagnosis using multilayer neural network with Levenberg Marquardt training algorithm," J. Med. Syst., vol. 35, no. 3, pp. 433-436, 2011, doi: 10.1007/s10916-009-9378-2.
13. E. Dogantekin, A. Dogantekin, and D. Avci, "Automatic hepatitis diagnosis system based on Linear Discriminant Analysis and Adaptive Network based on Fuzzy Inference System," Expert Syst. Appl., vol. 36, no. 8, pp. 11282-11286, 2009, doi: 10.1016/j.eswa.2009.03.021.
14. W. Ahmad et al., "Intelligent hepatitis diagnosis using adaptive neuro-fuzzy inference system and information gain method," Soft Comput., vol. 23, no. 21, pp. 10931-10938, 2019, doi: 10.1007/s00500-018-3643-6.
15. J. S. Sartakhti, M. H. Zangooei, and K. Mozafari, "Hepatitis disease diagnosis using a novel hybrid method based on support vector machine and simulated annealing (SVM-SA)," Comput. Methods Programs Biomed., vol. 108, no. 2, pp. 570-579, 2012, doi: 10.1016/j.cmpb.2011.08.003.
16. G. Sahebi, P. Movahedi, M. Ebrahimi, T. Pahikkala, J. Plosila, and H. Tenhunen, "GeFeS: A generalized wrapper feature selection approach for optimizing classification performance," Comput. Biol. Med., vol. 125, no. August, p. 103974, 2020, doi: 10.1016/j.compbiomed.2020.103974.
9_
17. M. Sinan Basarslan, B. H., and i.Yucedag, "Fuzzy Logic and Correlation-Based Hybrid Classification on Hepatitis Disease Data Set," in The International Conference on Artificial Intelligence and Applied Mathematics in Engineering, 2019, pp. 787-800, doi: 10.1007/978-3-030-36178-5.
18. G. V. Nivaan and A. W. R. Emanuel, "Analytic Predictive of Hepatitis using the Regression Logic Algorithm," 2020 3rd Int. Semin. Res. Inf. Technol. Intell. Syst. ISRITI 2020, pp. 106-110, 2020, doi: 10.1109/ISRITI51436.2020.9315365.
19. N. El Aboudi and L. Benhlima, "A new approach based on PCA and CE-SVM for hepatitis diagnosis," Lect. Notes Electr. Eng., vol. 381, pp. 91-99, 2016, doi: 10.1007/978-3-319-30298-0_10.
20. R. D. Prayogo and N. Ikhsan, "Attribute Selection Effect on Tree-Based Classifiers for Letter Recognition," 2020 Int. Conf. Data Sci. Artif. Intell. Bus. Anal. DATABIA 2020 - Proc., pp. 13-18, 2020, doi: 10.1109/DATABIA50434.2020.9190393.
21. J. Shlens, "A Tutorial on Principal Component Analysis," 2014, [Online]. Available: http://arxiv.org/abs/1404.1100.
22. S. Wold, K. Esbensen, and P. Geladi, "Principal component analysis," Chemom. Intell. Lab. Syst., vol. 2, no. 1-3, pp. 37-52, 1987, doi: 10.1016/B978-0-12-824447-0.00012-1.
23. D. Karaboga and E. Kaya, "Adaptive network based fuzzy inference system (ANFIS) training approaches: a comprehensive survey," Artif. Intell. Rev., vol. 52, no. 4, pp. 2263-2293, 2019, doi: 10.1007/s10462-017-9610-2.
24. J. S. R. Jang, "ANFIS: Adaptive-Network-Based Fuzzy Inference System," IEEE Trans. Syst. Man Cybern., vol. 23, no. 3, pp. 665-685, 1993, doi: 10.1109/21.256541.
25. M. Deif, R. Hammam, and A. Solyman, "Adaptive Neuro-Fuzzy Inference System (ANFIS) for Rapid Diagnosis of COVID-19 Cases Based on Routine Blood Tests," Int. J. Intell. Eng. Syst., vol. 14, no. 2, pp. 178-189, 2021, doi: 10.22266/ijies2021.0430.16.
26. D. Çaliçir and E. Dogantekin, "A new intelligent hepatitis diagnosis system: PCA-LSSVM," Expert Syst. Appl., vol. 38, no. 8, pp. 10705-10708, 2011, doi: 10.1016/j.eswa.2011.01.014.
27. M. Neshat, M. Sargolzaei, A. Nadjaran Toosi, and A. Masoumi, "Hepatitis Disease Diagnosis Using Hybrid Case Based Reasoning and Particle Swarm Optimization," ISRN Artif. Intell., vol. 2012, pp. 1-6, 2012, doi: 10.5402/2012/609718.
28. K. B. Nahato, K. N. Harichandran, and K. Arputharaj, "Knowledge mining from clinical datasets using rough sets and backpropagation neural network," Comput. Math. Methods Med., vol. 2015, 2015, doi: 10.1155/2015/460189.
29. H. L. Chen, D. Y. Liu, B. Yang, J. Liu, and G. Wang, "A new hybrid method based on local fisher discriminant analysis and support vector machines for hepatitis disease diagnosis," Expert Syst. Appl., vol. 38, no. 9, pp. 11796-11803, 2011, doi: 10.1016/j.eswa.2011.03.066.
30. A. Alamsyah and T. Fadila, "Increased accuracy of prediction hepatitis disease using the application of principal component analysis on a support vector machine," J. Phys. Conf. Ser., vol. 1968, no. 1, pp. 0-6, 2021, doi: 10.1088/1742-6596/1968/1/012016.