Diabetes is a chronic disorder caused by either inadequate insulin production by the pancreas or inadequate insulin absorption by the body. Many machine learning approaches handle a wide range of chronic conditions and keep track of patient health data. The analysis of medical data from various angles and the creation of knowledge from it can be accomplished using a variety of machine learning techniques. Creating new features by combining two or more features can provide more insights for health-related data. It aids in revealing a data set's hidden relationships. This work implements LR, RFECV-LR, and RFECV-SGDLR for comparison purposes and comes with the best suitable classification model. Further, this work suggests an IoT-based diabetes model that can also record information about their location, body temperature, and blood glucose levels and can help patients live healthier lifestyles by tracking their activities and diets.
Top2. Literature Survey
Minyechil Alehegn and Rahul Joshi 2017, Proposed Predictive algorithms such as KNN, Nave Bayes, Random Forest, and J48. Using PIDD data, an ensemble hybrid model was constructed by integrating any approaches and procedures into one which gives better performance (Alehegn, 2019). Deepti Sisodia, Dilip Singh Sisodia 2018 In this experiment, three ML classification algorithms, Decision Tree, SVM, and Naive Bayes are used to identify diabetic conditions. (PIDD) is utilized in experiments. The performance of matrices is just a handful to evaluate how well the three algorithms perform. Their results show that, when compared to other algorithms, Naive Bayes performs the best (Sisodia, 2018). KM Jyoti Rani 2020 proposed a diabetic prediction model that makes use of KNN, LR,Random Forest, SVM, and Decision Tree. Their experimental findings determine the suitability of the suggested system, with greater accuracy, reached utilizing the Decision Tree method (Jyothi Rani, n.d.). Ram D. Joshi and Chandra K. Dhakal 2021 recommended using a decision tree and logistic regression model to treat type 2 with the help of PIDD. According to their findings, the five most important predictors of type 2 diabetes are age, BMI, pregnancy, the function of the diabetic lineage, and glucose. The ten-node tree identifies significant predictors, while the six fold classification tree identifies key elements (Joshi & Dhakal, 2021). However, Prediction might be challenging for doctors to forecast diabetes. If machine learning is utilized, the patterns can be quickly found so that diabetes can be diagnosed early.