The Internet of Things (IoT) and machine learning (ML) are two rapidly evolving technologies that have profoundly impacted the healthcare section [1]. The convergence of these technologies has given rise to innovative and smart solutions in the form of IoT-based human-centered healthcare applications utilizing ML [2]. IoT-based healthcare applications for portable and remote monitoring with a focus on human-centricity are needed by those who live in underdeveloped or remote areas. Diseases like high blood pressure, heart disease, and diabetes are long-lasting and require continuous monitoring [3,4].

In IoT-based applications, sensors enable real-time data collection from diabetic patients, providing a continuous stream of vital signs and other health-related information. This constant monitoring ensures a comprehensive view of the patient’s health status. This vast amount of data is processed via ML to identify patterns, anomalies, and trends, assisting in personalized diagnosis and treatment plans. It can predict disease conditions and evaluate factors, on the basis of which risk assessments can be made for personalized interventions [5]. ML algorithms heavily rely on the quality and quantity of data. Existing data on diabetic patients are for patients who are demographically different and have different levels of health-service challenges as compared to local-area patients. Thus, inherent biases exist due to diverse demographic and environmental factors that affect human health. Inadequate or biased data may lead to incorrect predictions and recommendations [4,6].

Diabetes is a chronic disease or group of metabolic disorders. Blood glucose levels in the body are elevated, which is either due to a lack of insulin production or high sugar levels in the circulatory system [18]. This is due to malfunction of the pancreatic beta cells. It affects different parts of the body, and risks include pancreatic disease, heart disease, hypertension, kidney failure, pancreatic problems, nerve damage, foot problems, ketoacidosis, visual impairment, and many more. Eye problems, waterfalls, and glaucoma are also potential risks. There are various reasons behind the causes, such as lifestyle, lack of activity, weight gain, smoking, high cholesterol (hyperlipidemia), high blood pressure (hyperglycemia), and others, which can be targeted to treat diabetes. It affects various age groups, from adolescence to adulthood. The pancreas is an organ located in the midriff area. Gangil et al. [15] proposed a strategy that uses an ML technique, namely, an SVM, for classification for diabetes analysis. The ML strategy for diabetes management is based on a high-dimensional therapeutic data set. The experiment proved that the vector machine could be used to effectively diagnose diabetes.

Jara et al. [10] discussed smart sensing technologies that provide foundational data collection capabilities for IoT applications. Boursianis et al. [2] described the varieties of smart monitoring applications that have revolutionized several services not only in healthcare but in many other domains, like agriculture, networks, traffic control, and education, and in emergencies like fire disasters [22]. Challoner et al. [16] described smart healthcare services and digital health diagnosis performed using IoT. Alturjman et al. [4] discussed the authentication, integrity, and privacy of data using WSN in IoT. Rghioui et al. considered [23] continuous monitoring of patients using an IoT embedded system that makes health predictions based on machine learning algorithms, where an SMO algorithm gave the best results. Its results were evaluated in terms of accuracy, precision, and sensitivity. The IoT plays a dynamic role in the integration of physical and virtual items, as stated by Kavithamani et al. [1], who also discussed a health monitoring system based on IoT. IoT is crucial for developing cutting-edge technologies like wireless sensor networks, AI, cloud computing, robotics, transportation, and healthcare systems. According to the authors of [1,24], healthcare systems have seen many improvements, but glucose monitoring systems are still expensive and have several other drawbacks.

Smart healthcare or e-health plays a vital role in treatment and smart health prediction in diabetes. With the integration of technology and data-driven approaches, healthcare providers can offer more personalized and efficient care to diabetes patients [25]. Islam et al. [26] introduced the development of a smart healthcare monitoring system in an IoT environment. Five sensors are used: a heartbeat sensor, a body temperature sensor, a room temperature sensor, a CO sensor, and a CO₂ sensor. The patient readings are conveyed to the doctor’s portal and the doctor will give recommendations accordingly. The hardware they used was an ESP32, a heartbeat sensor, a temperature sensor, a CO sensor, and a CO₂ sensor. The architecture involves the collection of data from the sensors that are received by the ESP32, and then these measurements are transferred to a web server which is connected to the user interface of medical staff. The implementation details involve actual readings compared with observed readings for their proposed system [18].

Rghioui et al. [23] worked on a smart architecture for diabetic patient monitoring using ML algorithms. They monitored blood sugar levels, temperature, and physical activity using portable sensors and collected data. They collected data day-wise, in the morning, afternoon, and evening, in addition to the no. of steps taken in a day. The data are classified with a number of ML algorithms and compared, and the performance results are transferred to the doctor for deciding on the patient’s health.

Chatrati et al. [28] presented a smart home health monitoring system for predicting type 2 diabetes and hypertension. For these two diseases, they collected readings of blood pressure and blood glucose levels. For the prediction of status, they used ML algorithms and then gave notifications of status. The traditional approach was used; no rules were defined in it. The status representation is in the form of a category they found.

Qureshi et al. [29] presented an accurate and dynamic predictive model for a smart M-Health system using machine learning which they used to collect data through mobile applications. They proposed a secure Android application and reliable data storage and then transferred data for further ML processing. By means of ML algorithms, they classified cardiovascular diseases according to the seriousness of the conditions. They split the data into four folds, and predictive model decision trees and an SVM are used for prediction analysis. The performance of the predictive model with the benchmark was obtained in the form of accuracy, precision, and sensitivity.

There are diverse healthcare systems for diseases like blood pressure, hypertension, and diabetes [25]. These diseases require long-run continuous monitoring; thus, there is a need for portable smart healthcare systems that gather sensitive data from the patient’s body and transfer it to the relevant physician.

Sarwar et al. [27] worked on a fire detection system using an adaptive neuro fuzzy inference system. True detection of likelihood of fire is a novelty of their work. The implementation was performed in MATLAB. This system uses smart sensing technology for fire detection.

Afreen et al. [33] worked on an IoT-based monitoring and notification system using predictive analysis via an artificial neural network. Fruits and vegetables play an important role in minimizing the impact of some diseases; thus, the work focused on cold storage. The factors analyzed by the monitoring system were gauge temperature, relative humidity, and other vital ambient environmental parameters, such as luminosity and the concentration of gases.

Shafi et al. [12] also utilized IoT and machine learning for an agricultural storage combustion system. Cotton storage has a number of limitations and challenges, i.e., heat due to microbial growth, exothermic and endothermic reactions in storage areas, and extreme weather conditions in storage areas. Monitoring and real-time sensing predictions are performed by an artificial network and machine learning to control sudden change and to avoid damage to the quality of cotton. Saifullah et al. [34] also introduced a smart sensing system for radiation monitoring and a warning system. Radiofrequency electromagnetic rays have a significant impact on the human body; thus, using machine learning, various levels of EMR intensities are analyzed.

Table 3 provides a comparative analysis of IoT systems that employ a layered architecture and sensing parameters for data collection.

System Name	No. of Parameters	Sensing Parameters Monitored	Health Condition Checked	Communication Protocol	No. of Layers	Technology Used
A personalized healthcare Monitoring system for diabetic… [25]	4	Glucose, heart rate, activity, temperature	Diabetes, general health	Bluetooth, Wi-Fi	5	Mobile app, wearables
A real-time health monitoring system… [35]	3	Heart rate, blood pressure, temperature	Cardiovascular	Wi-Fi	4	Wearables, IoT, Zephyr BT
IoT-based personal health care monitoring device for diabetic patients [36]	2	Glucose, ketones	Diabetes	Wi-Fi	3	Mobile app, cloud
Wearable IoT enabled real-time health monitoring system [30]	3	Heartbeat, temperature, blood pressure	Diabetes, cardiovascular disease, obesity	Wi-Fi	3	Mobile app, cloud
Proposed article	5	Blood glucose, temperature, heart rate, blood pressure, oxygen saturation	Diabetes, general health	Wi-Fi	6	Wearable sensor, cloud

There were three articles found on IoT-based human-centered healthcare systems. A comparative analysis of key features of these three contributions is shown in Table 4. The key features discussed for comparative analysis are the ML algorithms used in each article; the data collection approach is smart sensing for remote monitoring with wearable sensors. Healthcare services can be handled at a personalized level, but their level, medium or high, varies according to the article.

Feature	Comparative	Study Analysis		Proposed Work
Focus of study	Remote patient monitoring and diagnostics [22,26]	Chronic disease management [20]	Telemedicine and consultation [32]	Smart sensing and remote monitoring of diabetic patients
ML algorithms used	SVM, random forest, KNN	Deep learning (CNN, LSTM)	Decision trees, Naïve Bayes	Support vector algorithm, Gaussian NB, decision tree, random forest, Bernoulli NB
Data collection approach	Wearable devices, health sensors	IoT sensors, electronic health records	Remote monitoring devices, health wearables	Wearable health sensors (5 sensors, including a specially designed glucose sensor for diabetic patients)
Personalization level	High	Medium	High	High
Real-time alerts	Yes	Yes	Yes	Yes
Patient engagement strategy	Personalized recommendations, gamification	Health tracking, goal setting	Educational content, interactive interfaces	Beep alerts, graph readings are available
Scalability and accessibility	Scalable, cloud-based	Scalable, no cloud access	Scalable	Scalable, cloud access
Decision support system	No	No	No	Yes
Smart data processing	No	No	No	Yes (give priority to emergency data)