INTRODUCTION:

Arthur Samuel introduced the term “Machine Learning” in 1959, describing it as a field that enables computers to learn from data without being explicitly programmed ¹. In simple terms, it involves developing algorithms that use past data and experiences to make predictions or decisions.

Machine learning is a branch of artificial intelligence that uses factual techniques to train computers to improve their performance continually. It works by identifying pattern and relationships within data, helping systems make predictions or decisions automatically ².

Machine learning (ML) has progressed significantly over the previous two decades, from a research curiosity to a viable technology with numerous commercial uses. It is a bra of artificial intelligence (AI) that uses statistical approaches to fit models to data and identify significant patterns from huge, unstructured, and complex data sets³. It is a complete, multidisciplinary field having roots in statistics, mathematics, computer science, and cognitive analytics, among other fields. Data observations, such as examples, first-hand knowledge, or instructions, serve as the foundation of the machine learning process. It looks for patterns in the data and then draws inferences from the provided examples. The primary goal of machine learning (ML) is to enable computers to learn autonomously and adapt after retraining. To predict future outputs, the supervised ML method trains a model using past data on both inputs and outputs, whereas unsupervised ML investigates intrinsic structures and hidden patterns in input data^{4, 5}.

The healthcare business has recently seen a significant impact from machine learning (ML) technologies. The use of machine learning techniques in healthcare can lead to breakthroughs such as more precise prediction models, innovative treatment approaches, and clinical decision support systems (CDSS), pharmaceutical discovery, and reductions in healthcare expenditures ^[6]. Recent practical applications of machine learning in healthcare have been made possible by the collecting of daily healthcare data and advancements in big data processing. Different ML techniques can be applied to these datasets, which can be organized or unstructured, to improve healthcare outcomes. Several ML methods, including linear regression (LR), support vector machine (SVM), random forest (RF), decision tree (DT), K-nearest neighbour (KNN), deep learning (DL), artificial neural network (ANN), and boosting algorithms, are commonly employed to forecast diseases ^7,8. With the advancement of big data analytics, increasing attention has been given to predicting diseases using data-driven approaches. Instead of relying on manually selected features, recent studies have focused on automatically identifying key characteristics from vast datasets to enhance the accuracy of risk classification ^9,10. However, those existing work mostly considered structured data. Furthermore, diseases can vary greatly across different regions, mainly due to differences in climate, lifestyle, and living habits. Therefore, the goal of this systematic review is to present a comprehensive overview of previous studies on predictive models for various communicable diseases. This review places particular emphasis on summarizing comparative data in tabular form, allowing readers to easily understand the datasets used, key findings, outcomes, and the main factors that contributed to improving the accuracy of the proposed models.

OBJECTIVES:

· To apply machine learning techniques for disease prediction.

· To pre-process and analyse healthcare data effectively.

· To evaluate and compare different machine learning models for accuracy.

· To build a simple interface for predicting disease risk.

· To assist in early diagnosis and preventive healthcare.

MACHINE LEARNING MODEL:

Machine learning is the automated learning of a computer system or machine from provided data, often known as raw data or datasets. This method normally consists of two primary parts:

1. Training of model.

2. Decision making or Testing of model.

We use two types of datasets: training and test data. To train the model, training data is pre-processed and features are extracted. Training of the model. The trained model is used to predict outcomes from unknown datasets, often known as test data. Model correctness may be assessed by making decisions based on previously known test data results¹¹.

Fig -1: Training and testing of machine learning model

Classification of Machine Learning Algorithms

Machine learning methods are broadly classified into three learning methods:

A. Supervised

B. Unsupervised

C. Reinforcement learning.

Fig -2: classification of machine learning algorithms.

A. Supervised Learning:

This algorithm requires supervision to make predictions or decisions. The input dataset is separated into training and test sets. During guided learning, the training data has already been allocated goal or output values for model training. Supervised learning is often used to solve classification and regression issues. Classification is utilized for discrete-value problems, whereas regression is used for continuous-value decisions¹².

Fig-3: Classification of Supervised Learning algorithms

Support vector machine (SVM):

This is the most prevalent approach to categorization difficulties. It operates by calculating margins. This strategy employs hyperplanes or a series of hyperplanes for separation. High-dimensional characteristics help distinguish across object types. SVM was initially limited to two-class or binary-classification issues. However, it was later expanded to answer multi-class problems using hyperplanes ¹³.

Discriminant Analysis:

To address the issue of dimensionality, consider reducing the number of data dimensions. Increased dimensions lead to a positive increase in space volume. The given dataset is sparse, leading to issues with statistical calculations. Discriminant analysis clusters items based on comparable traits to categorize them into different categories. This is an excellent example of a dimensionality reduction strategy. Popular discriminatory techniques rely on dimensional reduction are:

· Multidimensional Scaling (MDS)

· Linear Discriminant Analysis (LDA)

· Mixture Discriminant Analysis (MDA)

· Quadratic Discriminant Analysis (QDA)

· Flexible Discriminant Analysis (FDA) ¹⁴.

Naïve Bayes:

Naïve Bayes is a classification-based machine learning strategy that uses conditional probability to identify an object's membership. Whether a given class exists or not. A Bayesian network, also known as a tree network, is constructed based on the likelihood of belonging to a specific class ^[15].

Nearest Neighbour:

Nearest neighbour, also known as K-nearest neighbour (KNN), is an instance-based method. This technique makes predictions based on test results. Identify the item with the most comparable feature to those in training datasets. Larger training datasets lead to improved accuracy ^[16].

Ensembles Methods:

Until the 1990s, individual learning methods were widely used. However, in recent years, there has been an increase in the usage of ensemble learning approaches. A holistic approach to learning yields superior results compared to specific algorithms. The two most common ensemble approaches are:

a. Boosting

b. Bagging ¹⁷

Decision trees

Learning in decision trees includes predicting a model that maps an object's observations and conclusions to its goal value. Models of decision trees that Classification tree models are those with defined goal values. The decision tree's interior nodes represent object features, whereas leaf nodes represent class labels ¹⁸.

Some of the important decision tree methods are:

· Classification and Regression Tree (CART)

· Iterative Dichotomiser 3 (ID3)

· C4.5 and C5.0

· Chi-squared Automatic Interaction Detection (CHAID)

· Decision Stump

B. Unsupervised Learning:

Unsupervised learning uses inference to identify hidden patterns in unlabelled datasets. Here the model is trained. Unlabelled data is analyzed to make decisions for available test data. Unsupervised learning focuses on clustering and dimensional reduction challenges.

The most popular unsupervised algorithms are K-means clustering, Principal component analysis (PCA), Hierarchical model, Hidden Markov model, neural network, etc.

Fig-4: Unsupervised Learning Mechanism

K-means clustering:

The clustering algorithm divides data into groups based on similar features, with each group containing data with less distinguishable features within a certain distance. The K-means model uses a centroid-based approach with K-clusters, with the mean value located in the center of each cluster.

Principal component analysis (PCA):

PCA is a quick and efficient methodology compared to other methods. PCA reduces the dimensionality of data from high to low levels. Although dimensional data may be transformed linearly, it has limitations. This approach prioritizes key features over less relevant ones. This involves maintaining unique traits while removing comparable ones to minimize data dimensionality.

Neural networks:

Neural networks may also be utilized for unsupervised learning. The neural model has no clue about the intended output. This network classifies data. Classification is based on parameter similarity in data. The unsupervised neural network classifies things based on their input parameters¹⁹.

C. Reinforcement Learning:

This form of learning is commonly utilized in robotics, gaming, navigation, and other applications. This is dynamic learning. This strategy provides an atmosphere to accomplish a specific outcome without clearly indicating how close the aim is. The reinforcement model is a learning methodology that uses the try-and-hit method to pick the action that yields the highest positive outcome (closer to the aim). It is a combination of supervised and unsupervised learning models. The system starts by doing random acts and then delivers positive feedback to rectify them.

The reinforcement learning model relies on two functions: trial and error discovery and delayed outcomes. The reinforced model: the agent develops behaviour depending on Input i comes from the input function I, whereas state transition r comes from the current state (s). Behaviour B triggers action a, which modifies the environment²⁰.

The reinforcement neuron model relies on feedback to determine if the network's actions are corrective. If rectification is positive, the network makes a decision. Taking a good forward path might lead to a more accurate outcome. It should be noted that the network has no prior information about the target.

Fig-5: Reinforcement Learning Mechanism

CONCLUSION:

In this paper we reviewed various machine learning models that are most popular. Nowadays, people rely on machine learning-based intelligent systems for various purposes, both directly and indirectly. This study gives insight. This article provides an overview of popular machine learning models and their potential applications, as well as advancements in these methods. This information of relative performance can be used to aid researchers in the selection of an appropriate supervised machine learning algorithm for their studies.

REFERENCES:

1. Javed Azmi et.al. (July 2022) A systematic review on machine learning approaches for cardiovascular disease prediction using medical big data. Medical Engineering & Physics Volume 105.

2. Saiesh Jadhav et.al. (2019) Disease Prediction by Machine Learning from Healthcare Communities. IJSRST. Volume 6, Issue 3. Page No: 29-35.

3. Davenport T, Kalakota R (2019) Digital Technology The potential for artificial intelligence in healthcare.

4. Horvitz E, Mulligan D (2015) Data, privacy, and the greater good. Science 349(6245):253–255.

5. Allenbrand C (2024) Supervised and unsupervised learning models for pharmaceutical drug rating and classifica tion using consumer generated reviews. Healthc Anal 5:100288.

6. Devi MK et al (2022) Design and implementation of advanced machine learning management and its impact on better healthcare services: a multiple regression analysis approach (MRAA). Comput Math Methods Med.

7. Binson VA, Thomas S, Subramoniam M, Arun J, Naveen S, Madhu S (2024) A review of machine learning algorithms for biomedical applications. Ann Biomed Eng 52(5):1159–1183.

8. Osisanwo FY, Akinsola JET, Awodele O, Hinmikaiye JO, Olakanmi O, Akinjobi J (2017) Supervised machine learning algorithms: classification and comparison. Int J Comput Trends Technol 48(3):128–138.

9. B. Qian, X. Wang, N. Cao, H. Li, and Y.-G. Jiang, A relative similarity based method for interactive patient risk prediction, Data Mining Knowl. Discovery, vol. 29, no. 4, pp. 10701093, 2015.

A. Singh, G. Nadkarni, O. Gottesman, S. B. Ellis, E. P. Bottinger, and J. V. Guttag, Incorporating temporal EHR data in predictive models for risk strati cation of renal function deterioration, J. Biomed. Inform. vol. 53, pp. 220228, Feb. 2015.

10. Darpan Pandey, Kamal Niwaria, Bharti Chourasia. Machine Learning Algorithms: A Review. International Research Journal of Engineering and Technology (IRJET). Volume: 06 Issue: 02 | Feb 2019.

11. S.B. Kotsiantis, “Supervised Machine Learning: A Review of Classification Techniques”, Informatica 31 (2007) 249-268.

D. Meyer, “Support Vector Machines – The Interface to libsvm in package 1071”, August 2015.

12. Y. Lee, Y. Lin, and G. Wahba, “Multicategory support vector machines: Theory and application to the classification of microarray data and satellite radiance data,” Journal of the American Statistical Association, 99(465):67–81, March 2004.

E. Lowd, P. Domingos, “Naïve Bayes Models for Probability Estimation,” ICML '05 Proceedings of the 22nd international conference on Machine learning, 2005.

13. J. M. Keller, M. R. Gray, J. A. Givens Jr., “A Fuzzy K-Nearest Neighbor Algorithm”, IEEE Transactions on Systems, Man and Cybernetics, Vol. SMC-15, No. 4, August 1985.

14. Z.H.Zhou, “Ensemble Learning”, National Key Laboratory for Novel Software Technology, Nanjing University, Nanjing, China, 2013.

15. L. Rokach and O. Maimon, "Top-down induction of decision trees classifiers - a survey," IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews), vol. 35, no. 4, pp. 476-487, Nov. 2005.

16. K. Alsabati, S. Ranaka, V. Singh, “An efficient k-means clustering algorithm”, Electrical Engineering and Computer Science, 1997.

17. R.S. Sutton, “Introduction: The Challenge of Reinforcement Learning”, Machine Learning, 8, Page 225-227, Kluwer Academic Publishers, Boston, 1992.

Received on 10.11.2025 Revised on 04.02.2026

Accepted on 06.03.2026 Published on 24.06.2026

Available online from June 30, 2026

International Journal of Technology. 2026; 16(1):59-64.

DOI: 10.52711/2231-3915.2026.00007

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. Creative Commons License.