Challenges in Information, Communication and Computing Technology – V. Sharmila et al.

(Eds)
© 2025 The Author(s), London, 978-1-032-90166-4
Open Access: www.taylorandfrancis.com, CC BY-NC-ND 4.0 license

Smart health monitoring system using IoT and machine learning

S. Thangam, M. Gurupriya, A. Adithya Vardhan, Maganti Akshita Uma Sai &

Yelleti Manikanta Praneeth Kumar
Department of CSE, Amrita School of Computing, Amrita Vishwa Vidyapeetham, Bangalore, India

ABSTRACT: The fundamental element of people’s needs is health. Humans face a haul of
surprising death and plenty of diseases because of varied diseases resulting from a lack of
treatment to patients at the right time. By seamlessly integrating the Internet of Things (IoT),
Machine Learning, and Shapley Additive Explanations (SHAP), this study presents a revolu-
tionary paradigm for health monitoring. The system collects real-time health data via IoT
devices. Machine learning approaches are used for comprehensive health pattern analysis, and
SHAP values aid in assessing interpretability and feature importance. The Random Forest
model achieved an excellent accuracy of 60% in the experiments, demonstrating a substantial
breakthrough. This study demonstrates the efficacy of combining IoT with machine learning,
highlighting its potential for developing health monitoring systems. The method not only pro-
duces accurate forecasts, but it also provides essential insights into the critical elements impacting
health outcomes, hence helping to the growth of tailored and data-driven healthcare solutions.

Keywords: Internet of Things (IOT), Machine Learning, Temperature Sensor, Pulse Sensor,
WIFI Module, Arduino, ThingSpeak

1 INTRODUCTION

The Internet of Medical Things (IoMT) revolutionizes healthcare by linking objects with
sensors and information technologies to enhance patient care, research, and monitoring.
Remote monitoring of chronic diseases using IoT and wireless solutions promises to improve
billions of lives. IoMT communicates via sensors, gateways, and networks, crucial for dis-
ease management and healthcare quality. By 2025, a trillion-dollar investment will drive the
adoption of wearable devices, networked sensors, and data analytics, boosting accessibility,
personalized care, and health promotion. Despite challenges like under-resourced hospitals
and disparities in care, IoMT’s continuous data collection and analysis can transform
healthcare delivery and efficiency.

2 LITERATURE REVIEW

An overview of remote patient health monitoring system [1]. It discusses the use of various
sensors and IoT devices to monitor the health parameters of patients. The system is designed
to capture data and transmit it over the internet for analysis, allowing for remote monitoring
of patients’ health. A real-time patient monitoring system working with Arduino and
Raspberry Pi boards, temperature sensors and pulse oximeters [3]. Data is passed to author-
ized devices through a cloud server, demonstrating aspects such as Raspberry Pi, Arduino Uno
and MQTT protocol. Results illustrate that successful data is conveniently presented on
smartphones and laptops with MQTT Lens. Research work [4] presented a detailed study of
patient health monitoring system with the help of IoT technology and cloud-based infra-
structure. The data is collected from health users through IoT devices then this information

518 DOI: 10.1201/9781003559085-90

This chapter has been made available under a CC BY-NC-ND 4.0 license
gets transmitted processed and stored in the cloud for continuous monitoring and analysis.
The process of building a healthcare monitoring system based on IoT [5], cloud computation
and container-based virtualization by using edge devices. The main components of the system
included Internet of Things IoT for data collection, cloud-based infrastructure for storage and
processing data, container based virtual environments on edge devices to optimize deployment
efficiency. It makes the system more flexible and scalable when using containerization in edge
devices. This paper overall discusses the design, implementation, and potential advantages of
this integrated healthcare monitoring system, showcasing advancements in the field. Yang et.
al. [6] created a wearable system for ECG monitoring within the scope of smart healthcare.
This system is made up with IoT and cloud-based technologies. The wearable device is pre-
dicted to record ECG. The wearable device is expected to capture ECG data, which is then
transmitted through IoT connectivity to a cloud infrastructure for storage and analysis. The
article [7] discusses the key components of architecture designed for smart healthcare systems
that incorporates Internet of Things (IoT) in healthcare, and the benefits of incorporating IoT
awareness. Their aim is to provide a foundation for developing intelligent and interconnected
healthcare solutions. The work [8] proposed a method for ubiquitous data access in an Internet
of Things (IoT)-based information system specifically designed for emergency medical ser-
vices. The implementation [9] of an IoT-based smart nursing system for patients who are
bedridden. In [10] presents a portable health monitoring system that helps patients have
remote access to the healthcare providers and monitor their wellbeing, especially in COVID-19
time. It is made up of Raspberry Pi, Node-Red, ThingSpeak and also sensors which include
LM35 temperature sensor, AD8232 ECG Sensor and MAX 10 pulse oximeter to measure key
health parameters which allows doctors or guardians to monitor patients remotely. The data
gathered by the sensors is then sent to a cloud-based platform for monitoring in real time and
can also be used when offline. Article [11] proposes a model for a smart healthcare system
leveraging IoT-cloud technology. It outlines the use of sensors connected to a Raspberry Pi to
monitor patients’ symptoms, with data transmission to a cloud server for access by healthcare
professionals. The challenges of implementing IoT in healthcare are discussed, and a com-
parison between the use of docker containers and virtual machines in the proposed model is
provided. Research work [12] discuses about deploying IoT based real-time pulse and tem-
perature monitoring systems in healthcare. Utilization of sensors, a Raspberry Pi micro-
processor, and a cloud platform for continuous monitoring on patients’ vital statistics and live
data presentation and also Proteus for simulation purpose was used in [13].

3 METHODOLOGY

3.1 Proposed method

Figure 1 shows the proposed system which includes Arduino Uno, LM35 Temperature
Sensor, Pulse Sensor, AD8232 and ADXL 345. The patient’s heart is the Arduino Uno. The
sensors capture the data which is then relayed to Arduino Uno. Arduino Uno processes the
data and sends it to the ThingSpeak cloud platform. Now, the data or alerts can be visua-
lized using ThingSpeak platform in case specific conditions are met in monitoring system.

Figure 1. Architecture of the health monitoring system.

519
Table 1. The components used for the IOT-based health monitoring system.

Model Specifications

Node-MCU 32 pin Wi-Fi module with built-in Wi-Fi, enabling wireless connectivity and easy
ESP8266 integration for various Arduino projects, home automation, and sensor applica-
tions. It can be programmed using the Arduino IDE
LM35 5v-3 pin model is a temperature sensor widely used for accurate temperature
measurement in electronic circuits.
ADXL345 5v-5 pin model is utilized for fall detection in wearable devices and elderly care
systems by sensing sudden changes in acceleration, triggering alerts or safety
mechanisms for potential fall incidents
Pulse sensor 5v-5 pin model is employed for real-time heart rate monitoring providing a non-
invasive way to measure and track a person’s pulse
Ad8232 3.3v-5 pin model is an ECG sensor used to measure and monitor electrocardiogram
signals, detecting heart activity for health monitoring devices and medical
diagnostics

The components used for the IOT-based health monitoring system is shown in Table 1.
Figure 2 represents the results obtained from dashboard displaying four different health
monitoring systems. Every system has two graphs. The bottom graph depicts accelerometer
data, while the top graph depicts a heartbeat plot over time. The bottom right system shows
irregular accelerometer data, which may imply movement or activity, while the top left
system looks to have a fast heart rate. The other two systems had lower heart rates and less
variability in accelerometer data. In general, two of the four systems exhibit elevated heart
rates or activity, while the other two are quite tranquil. For each system, there is a top graph
displaying heart rate over time and a bottom graph displaying accelerometer data, which
most likely depicts motor activity. Framework This system’s heart beat appears to be high
and fluctuating regularly. Additionally, there are occasional fluctuations in the accel-
erometer data that indicate motion or activity. System 2 appears to be steadier, with a lower
heart rate than System 1. The accelerometer data also shows less fluctuation, indicating that
there is less movement or activity. System 3 has a modest and stable heart rate, similar to
System 2. The accelerometer data shows little variations. System 4: This system’s heart rate
is high and fluctuates regularly, similar to System 1. The accelerometer data also reveals
occasional variations, which indicate movement or activity. Overall, the dashboard indicates
that two of the four systems have elevated heart rates or activity, whilst the other two are
generally tranquil.

4 RESULTS AND DISCUSSION

Figure 2. Thing speak result.

520
4.1 Machine learning
Thirteen machine learning models were used, with regression models being Linear regres-
sion, ridge regression, lasso regression, decision tree regressor, random forest regressor, k-
neigbours regressor, support vector regressor and gradient boosting regressor. For regression
models the performance metrics are MSE, RMSE, MAE and MAPE. Classification models
are decision tree classifier, random forest classifier, kneigbours classifier, support vector
classifier and gradient boosting classifier and their performance metrics is accuracy, preci-
sion, recall and F1-Score.

Table 2. Regression models and their corresponding performance metrics.

Model Names MSE RMSE MAE MAPE

Linear Regression 684.38 26.16 23.31 31.31%

Ridge Regressor 684.37 26.16 23.31 31.31%
Lasso Regressor 684.75 26.17 23.34 31.36%
Decision Tree Regressor 798.10 28.25 25.09 33.01%
Random Forest Regressor 785.72 28.03 24.89 32.75%
K Neighbors Regressor 874.02 29.56 25.08 33.83%
SVR 700.21 26.46 23.77 32.30%
Gradient Boosting Regressor 801.82 28.32 25.09 32.97%

Table 2 shows that the Pulse Rate predictions made by the linear regression model have an
average error of 26.16. As anticipated, the results resemble those of linear regression. In the
event that the features exhibit multicollinearity, ridge regularization may be beneficial. MSE
and RMSE are marginally higher than in linear regression. It’s possible that Lasso is pena-
lizing some features more severely. When compared to linear models, the MSE and RMSE
are higher, indicating overfitting of the training set. The Regressor for Random Forest. The
advantage of an ensemble approach is demonstrated by the results, which are marginally
better than a single Decision Tree. Greater errors in comparison to linear models suggest
that the dataset may not be a good fit for the straightforward k-nearest neighbor method.
Linear models can be compared to MSE and RMSE.SVR can be sensitive to hyperpara-
meter tuning. Results are similar to the Decision Tree Regressor. Gradient Boosting might
require further tuning to improve performance (Figure 3).

Table 3. Classification models and their corresponding performance metrics.

Model Name Accuracy Precision Recall F1 Score

Decision Tree Classifier 57.38% 57.82% 57.38% 57.54%

Random Forest Classifier 60.66% 62.94% 60.66% 61.28%
K Neighbors Classifier 39.34% 40.53% 39.34% 39.67%
SVC 44.26% 50.94% 44.26% 35.44%
Gradient Boosting Classifier 49.18% 52.97% 49.18% 48.82%

Figure 3. Roc Curve for the classification machine learning models.

521
 Table 3 shows that the Decision Tree Classifier is making predictions with an accuracy of
57.38%. Higher accuracy, precision, recall, and F1 score compared to the Decision Tree
Classifier. Random Forests outperform others; K-neighbors has lower accuracy. SVC and
Gradient Boosting are moderate but need tuning for better performance. Figure 4,
90.5,690.6,90.7,90.8 plot shows the impact of temperature and pulse rate on the output of the
prediction model. Figure 4 shows the summary Plot, dependency Plot, decision Plot and
waterfall Plot.

Figure 4. Summary plot, dependency plot, decision plot & waterfall plot.

5 CONCLUSION

This research presents a novel approach to health monitoring, integrating IoT, machine
learning, and Shapley values, with a focus on transparency and interpretability. The appli-
cation of a Random Forest machine learning model yielded a commendable accuracy of
60%, showcasing the system’s potential for robust health assessments. The incorporation of
Shapley values not only augmented the accuracy but also provided valuable insights into the
model’s decision logic, enhancing trust and understanding. Despite achieving a significant
milestone, there remains room for improvement, and future work will explore model
refinement, hyper parameter tuning, feature engineering, and the incorporation of additional
data sources to further enhance predictive capabilities.

REFERENCES

[1] Akram P. S., Ramesha M., Valiveti. S, Sohail S. and Rao K. T. S. S. (2021). IoT based remote patient
health monitoring system, 7th International Conference on Advanced Computing & Communication
Systems (ICACCS), Coimbatore, India, pp. 1519–1524, doi: 10.1109/ICACCS51430.2021.9441874.
[2] Dandu R. and T. S. (2023). IoT based single axis solar tracker, 2023 14th International Conference on
Computing Communication and Networking Technologies (ICCCNT), Delhi, India, pp. 1–5,
[3] Gundala J. R., Varsha Potluri S. S., Damle S. V. and Hashmi M. F. (2022). IoT & ML-based healthcare
monitoring system-review, IEEE International Symposium on Smart Electronic Systems (iSES),
Warangal, India, pp. 623–626, doi: 10.1109/iSES54909.2022.00137.
[4] Gurupriya M., Vasavi C. S., Gayatri Ramasamy and Karthikeyan B. (2023). Combinatorial bat opti-
mization for WSN with hoover index clustering, International Conference on Next Generation
Electronics (NEleX), Vellore, India, 2023, pp. 1–5, doi: 10.1109/NEleX59773.2023.10421682.
[5] Jaiswal K., Sobhanayak S., Mohanta B. K. and Jena D. (2017). IoT-cloud Based Framework for
Patient’s Data Collection in Smart Healthcare

522
 [6] Kamarozaman N. B and Awang A. H. (2021). IOT COVID-19 Portable health monitoring system
using raspberry pi, node-red and thing speak, IEEE Symposium on Wireless Technology & Applications
(ISWTA), Shah Alam, Malaysia, pp. 107–112, doi: 10.1109/ISWTA52208.2021.9587444.
[7] Krishna C. S. and Sampath N. (2017). Healthcare monitoring system based on IoT, 2nd International
Conference on Computational Systems and Information Technology for Sustainable Solution (CSITSS),
Bengaluru, India.
[8] Maneesha D., Ramesh. V., Pathinarupothi, Krishnan R, and Ekanath Srihari Rangan. (2018). Systems,
Methods, and Devices for Remote Health Monitoring and Management.
[9] Niranjan D. K. and Rakesh N. (2020). Real time analysis of air pollution prediction using IoT, Second
International Conference on Inventive Research in Computing Applications (ICIRCA), Coimbatore,
India, pp. 904–909.
[10] Pathinarupothi R. K., Durga P. and Rangan E. S. (2019). IoT-based smart edge for global health:
remote monitoring with severity detection and alerts transmission, in IEEE Internet of Things Journal,
vol. 6, no. 2, pp. 2449–2462
[11] Sumathi A. and Gurupriya M. (2022), Efficient routing and performance amelioration using hybrid
diffusion clustering scheme in heterogeneous wireless sensor network, International journal of commu-
nication systems 35(15), 1–29
[12] Thangam S. (2021). Smart agriculture and role of IOT, 2021 Third International Conference on Inventive
Research in Computing Applications (ICIRCA), Coimbatore, India, pp. 1–6, doi: 10.1109/
ICIRCA51532.2021.9545042.
[13] Vivek P. S, Rahul P. V. S, D. E. S and Yadav S. (2021). IoT based smart nursing system for bedridden
patients, 2nd International Conference on Smart Electronics and Communication (ICOSEC), Trichy,
India, pp. 1–5.
[14] Selvaperumal S., Sivagama Sundari M. S. and SivaSubramanian K. (2018). Asymmetrical cascaded H-
Bridge 41 level inverter with single PV system, 2018 IEEE International Conference on Computational
Intelligence and Computing Research (ICCIC), Madurai, India, 2018, pp. 1–7.
[15] Kumar U., Mishra S. and Dash K. (2023). An IoT and semi-supervised learning-based sensorless
technique for panel level solar photovoltaic array fault diagnosis, in IEEE Transactions on
Instrumentation and Measurement, vol. 72, pp. 1–12.

523