The great majority of diabetes patients in India provide a unique set of challenges, and the prospective availability of data may significantly present a unique opportunity for efficiently addressing these challenges. If all doctors use electronic medical records to obtain this data, India may have a great chance to become a leader in this field of study. In this endeavor, the necessary electronic devices are routinely used to collect patient data. Artificial intelligence would help identify upcoming problems and perhaps even assist in developing solutions that are especially geared to make dealing with them a possibility. The possibility of a diabetic patient having a problem might be fixed by using different kinds of machine learning algorithms, which would boost the success rate of therapy. Along with XGboost and support vector machines (SVM), random forest is a well-known technique for making this prediction and managing the therapy, similar to the decision tree. In comparison to other classifiers, tuned XGB classifier produces the best results with an accuracy of 91%.
Top2. Literature Survey
As everyone is aware, the nine other criteria that were most frequently used in earlier studies involving diabetes prediction models were also taken into account, including the patient's age, gender, and family history of diabetes. The consumption of alcohol, body mass index (BMI), occasional smoking status of individuals, expanding waistlines, and regular physical activity are just a few of the many factors that might affect diabetes. (Cho et al., 2021; Chung et al., 2023; Edlitz & Segal, 2022; Hahn et al., 2022; Jonas et al., 2021; Schwatka et al., 2021). FPG was calculated using glucose readings that were obtained after a fast of roughly 8 hours. They only considered FPG, even though there are three techniques to detect prediabetes. Only the parents and siblings have a history of diabetes in the family.
The amount of alcohol was calculated based on the number of glasses, regardless of the type of liquid, with the assumption that each glass would have roughly the same amount of alcohol (e.g., 8 g). Respondents who had never smoked or had quit smoking were included in the “currently smoking frequently” and “others” categories for smoking status. The respondents who responded with a value larger than “moderate” to the question, “How intensive is your everyday activity?,” were considered to be physically active.
The structure of genuine neurons, such as those present in a human being's brain, served as a model for the ANN artificial intelligence technology (Han et al., 2018). The method, which focuses on categorisation, is mostly used in medicine to discover hidden patterns in risk factors. Neural networks are believed to be more reliable future predictors when properly trained than more established methods like logistic regression. The optimum prediction model can now be found automatically thanks to recent advancements in ANN technology (Zhang et al., 2021; Zhang et al., 2023). Contrary to logistic regression, ANNs can detect complex nonlinear relationships between a number of factors and diseases, making them useful in systems that support medical judgments (Ulivieri et al., 2021; Yuan et al., 2023).
Sisodia and Sisodia (2018) offer a different strategy based on support vector machine (SVM) techniques for categorizing individuals with and without prevalent ailments. They illustrate the method for identifying patients with diabetes and pre-diabetes using a cross-sectional representative sample of the American population. If they do not receive treatment, prediabetic women with a history of gestational diabetes mellitus (GDM) are more likely to develop type 2 diabetes than women without a history of GDM.
Both metformin and significant lifestyle modifications can reduce risk. To predict risk and therapy response specifically for women with prior GDM, we developed a risk forecasting method. The prenatal hyperglycemic microenvironment as seen in GDM-complicated pregnancies may not merely represent but additionally contribute to the epidemic of type 2 diabetes mellitus (T2DM) (Zhu & Zhang, 2016). Researchers thoroughly examined the data that was available over the previous ten years in an effort to assess the current worldwide prevalence of GDM by nation and area.