Article Preview
TopIntroduction
In recent decades, artificial intelligence (AI) did a huge commitment to computer engineering and a significant number of its application areas. Machine learning (ML), by producing algorithms that learn decision rules and patterns through data, has recently received a lot of attention. Data mining methodologies can also be utilized to generate novel prediction models that, starting with pre-existing risk forecast calculators, can be merged with data from a particular clinical depository to correctly help disease control and affected person care (Dagliati et al., 2018).
Data mining is the practice in which useful and previously unknown patterns are extracted from a large database or data warehouse. Today, data mining plays an essential role in a wide range of industries, such as the health care, banking, and financial sectors, as well as the education sector. Diverse studies have been conducted on diabetes prediction using different algorithms. Healthcare organizations collect a large amount of data. When data mining techniques are utilized to create models that learn from observable data, new information is gained (Alam et al., 2019).
Diabetes is becoming more common across the world as a result of environmental and genetic causes. The numbers are quickly increasing due to a variety of causes, including poor diets, physical inactivity, and others. In 1980, there were 122 million diabetics around the world, and by 2019, there were 463 million. Around 700 million people are expected to have diabetes by 2045 (https://idf.org/aboutdiabetes/what-is-diabetes/facts-figures.html). Moreover, diabetes was directly responsible for approximately 1.6 million fatalities. Diabetes Mellitus is among the most prevalent chronic childhood disorders, affecting children, adolescents, and young adults. An inability to produce enough insulin to regulate blood sugar levels is referred to as diabetes. So, it is a hormonal disorder when the body fails to generate insulin causes improper sugar metabolism, resulting in elevated blood glucose levels in the body of a specific individual. Some of the visible features are intense appetite, thirst, and frequent urination. Certain risk variables, including body mass index (BMI), age, blood pressure, glucose levels, and other variables, all play a role in the disease's impact (Diwani & Sam, 2014; Lyngdoh et al., 2021; Maniruzzaman et al., 2020).
Gestational, prediabetes, type 2, and type 1, delimit the 04 diabetes’ kinds. The pancreas is not producing insulin in persons with type 1. In non-insulin-dependent diabetes, commonly known as type 2, the body produces enough insulin but cannot use it. During pregnancy, gestational diabetes occurs. When blood glucose’s level is higher than usual however not higher enough to diagnose, it is called prediabetes (Jayanthi et al., 2017).
Prediction of type 2 diabetes onset in its early stages is a hotly debated health topic. The diabetes risk score was the most convenient instrument for predicting diabetes at early stages of the malady. This strategy, on the other hand, relies on human decision-making. Computing models that predict diabetes risk can help healthcare providers make better decisions and assist patients in the self-management of the disease, which can reduce the disease's associated death rates. Since ML approaches perform well in predicting diabetes, they are gaining traction in the health profession. They can assist with identifying people with higher danger of being diabetic and who can profit from early detection, prevention, and treatment programs. This reduces the risk of human error when making important healthcare decisions. As a result, the health burden is reduced, and resources are better utilized (Alshammari et al., 2020).
Diabetes data is challenging to evaluate because of non normal, non linear, and complicated nature of medical data. Medical imaging and healthcare have been dominated by ML based systems. Moreover, ML can be utilized as feature selection approaches and also as classifiers, according to researchers. For precise diabetes risk stratification, it also aids in the correct diagnosis of diabetes (Maniruzzaman et al., 2020).