Predictive Maintenance in Healthcare IoT: A Machine Learning-based Approach

Abstract

The increasing adoption of Internet of Things (IoT) devices in healthcare has created new opportunities for optimising medical equipment maintenance and enhancing patient care. Predictive maintenance, enabled by machine learning techniques, has emerged as a promising strategy for improving the dependability and effectiveness of healthcare IoT systems. This research paper presents a thorough examination of the application of machine learning for predictive maintenance in the Internet of Things (IoT) for healthcare. Embedded Internet of Things (IoT) sensors in medical devices such as vital signs monitors, infusion pumps, and imaging systems are utilised to collect data for the proposed method. These sensors continuously collect data in real time regarding the performance, operating conditions, and environmental factors of the equipment. Advanced machine learning algorithms are applied to the collected data to identify patterns and trends indicative of equipment failures or degradation. The research investigates various machine learning techniques, such as supervised and unsupervised learning, to develop accurate predictive models for IoT maintenance in healthcare. The models are trained with historical data to determine the relationship between sensor readings and maintenance events. The models are then deployed to predict potential equipment failures and trigger proactive maintenance actions, thereby minimising downtime and ensuring uninterrupted patient care.

Full text

Predictive Maintenance in Healthcare IoT: A Machine Learning-based Approach
Section A-Research paper
ISSN 2063-5346
15909
Eur. Chem. Bull. 2023, 12(Special Issue 4), 15909-15922
Predictive Maintenance in Healthcare IoT: A Machine
Learning-based Approach
Mrs. Priya Talawar1, Dr. Hemlata Pant2, Mr.Sohail Sayyed3, Mrs.Summaiya Tamboli4,
Mr.Nazeer Shaik5, Mr Veeramani Ganesan6
1,3,4Assistant Professor, BCA (SCIENCE), Abeda Inamdar Senior College of Arts, Science
and Commerce, Pune, Maharashtra, India.
2Associate Professor, Department of Computer Science & Engineering (CSE), Ambalika
Institute of Management and Technology (AIMT), Lucknow, UP, India.
5Assistant Professor, Department of CSE, Srinivasa Ramanujan Institute of Technology-
Autonomous, Anantapur, Andhra Pradesh, India.
6Sr Mobile and OTT Engineer, CBS News, Paramount, New York City, New York.
Abstract
The increasing adoption of Internet of Things (IoT) devices in healthcare has created new
opportunities for optimising medical equipment maintenance and enhancing patient care.
Predictive maintenance, enabled by machine learning techniques, has emerged as a promising
strategy for improving the dependability and effectiveness of healthcare IoT systems. This
research paper presents a thorough examination of the application of machine learning for
predictive maintenance in the Internet of Things (IoT) for healthcare. Embedded Internet of
Things (IoT) sensors in medical devices such as vital signs monitors, infusion pumps, and
imaging systems are utilised to collect data for the proposed method. These sensors
continuously collect data in real time regarding the performance, operating conditions, and
environmental factors of the equipment. Advanced machine learning algorithms are applied
to the collected data to identify patterns and trends indicative of equipment failures or
degradation. The research investigates various machine learning techniques, such as
supervised and unsupervised learning, to develop accurate predictive models for IoT
maintenance in healthcare. The models are trained with historical data to determine the
relationship between sensor readings and maintenance events. The models are then deployed
to predict potential equipment failures and trigger proactive maintenance actions, thereby
minimising downtime and ensuring uninterrupted patient care.
Keywords: Machine Learning, Sensor, IoT, healthcare and predictive models
1. Introduction
The rapid adoption of Internet of Things (IoT) devices in the healthcare sector has completely
changed how medical equipment is monitored and managed. Vital sign monitors, infusion
pumps, and imaging systems are a few examples of IoT-enabled medical devices that gather

Predictive Maintenance in Healthcare IoT: A Machine Learning-based Approach
Section A-Research paper
ISSN 2063-5346
15910
Eur. Chem. Bull. 2023, 12(Special Issue 4), 15909-15922
and send a tonne of real-time data. The performance of the device, operational effectiveness,
and patient health are all improved by using this data. To guarantee continuous patient care,
these devices' dependability and accessibility are crucial. Repairs and replacements are
typically made after equipment fails in traditional maintenance procedures used in the
healthcare industry. Increased downtime, expensive repairs, and potential patient safety risks
are all consequences of this method. Predictive maintenance is a proactive approach that uses
data analytics and machine learning algorithms to anticipate equipment failures or
degradation in order to overcome these difficulties. In the context of healthcare IoT,
predictive maintenance entails examining sensor data gathered from medical devices to find
patterns and anomalies that could point to potential equipment problems. In order to
accurately predict future failures, machine learning algorithms are applied to this data to learn
the correlation between sensor readings and maintenance events. Healthcare providers can
schedule proactive maintenance tasks by anticipating maintenance requirements, which
reduces downtime and maximises the use of medical equipment. This research paper's goal is
to investigate the use of predictive maintenance based on machine learning in the IoT for
healthcare. Our goal is to research various machine learning methodologies and create
models that can accurately predict the need for maintenance on medical devices. We will also
talk about the difficulties and factors to take into account when implementing predictive
maintenance in healthcare settings. Numerous important advantages come from the
incorporation of predictive maintenance in healthcare IoT systems. First of all, it enables
medical professionals to be proactive by planning maintenance tasks for times when they are
not urgent, lessening interruptions to patient care. Second, by spotting potential problems
early and enabling prompt repairs or replacements, it increases the lifespan of medical
equipment. Thirdly, it reduces the need for emergency replacements and unscheduled repairs,
which helps to reduce costs. This research paper will examine various machine learning
algorithms, analyse the body of existing literature on predictive maintenance in healthcare
IoT, and propose a framework for implementing predictive maintenance in healthcare settings
in order to accomplish these goals. We'll also go over the implications of data security and
privacy, as well as how to integrate preventive maintenance programmes with the current
healthcare system. The study's findings will add to the body of knowledge on preventative
maintenance in the IoT for healthcare and offer useful information for healthcare providers,
medical professionals, and equipment manufacturers. The adoption of machine learning-
based predictive maintenance in healthcare IoT systems will ultimately improve operational
effectiveness, improve patient care, and spur advancements in the healthcare sector.
2. Literature Survey
Numerous studies have been carried out to investigate the potential and effectiveness of
predictive maintenance in the IoT for healthcare sector. The purpose of this literature survey
is to present an overview of the body of research in this area and to highlight significant
discoveries and developments. A survey of preventative maintenance methods used in
healthcare IoT is presented in this survey paper. It provides a thorough analysis of numerous
preventative maintenance methods used in healthcare IoT. It discusses the use of various

Predictive Maintenance in Healthcare IoT: A Machine Learning-based Approach
Section A-Research paper
ISSN 2063-5346
15911
Eur. Chem. Bull. 2023, 12(Special Issue 4), 15909-15922
machine learning algorithms, including support vector machines, decision trees, and neural
networks, to predict equipment failures. The study also looks at the opportunities and
difficulties of implementing predictive maintenance in IoT systems for healthcare. Healthcare
IoT anomaly detection: By spotting anomalous patterns or outliers that could be signs of
impending equipment failure, anomaly detection techniques are essential for predictive
maintenance. In order to understand anomaly detection methods for healthcare IoT predictive
maintenance, this research focuses on statistical approaches, clustering algorithms, and deep
learning-based techniques. The study discusses the applicability of these techniques in actual
healthcare scenarios as well as their effectiveness. Predictive maintenance integration with
hospital management systems is examined in this study. Predictive maintenance solutions are
integrated with current hospital management systems. It investigates the compatibility of
asset management programmes, scheduling platforms, and electronic health records with
predictive maintenance software. The study highlights the advantages of seamless integration,
including enhanced data-driven decision-making, automated work order generation, and
optimal resource allocation. IoT predictive maintenance in healthcare: privacy and security
issues Given the sensitivity of healthcare data, privacy and security issues are of utmost
importance when implementing predictive maintenance. This study investigates the privacy
dangers of gathering and using patient health information for predictive maintenance. It
discusses data anonymization methods, access control systems, and encryption methods to
protect patient privacy while facilitating efficient predictive maintenance strategies.
Healthcare IoT proactive maintenance strategies: This study focuses on proactive
maintenance methods that go beyond failure prediction. In this study, predictive models that
continuously learn and adjust to changing equipment conditions are examined in relation to
adaptive maintenance approaches. The study investigates the use of dynamic scheduling
methods and reinforcement learning algorithms to maximise maintenance activities and
reduce interruptions to patient care. Case studies on the use of predictive maintenance in IoT
in healthcare To assess the practical application of predictive maintenance in healthcare IoT
environments, several case studies have been carried out. These studies look at the gains
made, such as increased patient satisfaction, lower maintenance costs, and increased
equipment uptime. They offer insights into real-world issues, approaches to implementation,
and takeaways from applying predictive maintenance techniques in healthcare settings. The
literature review, taken as a whole, emphasises the growing interest in predictive maintenance
in healthcare IoT and its potential to completely transform maintenance procedures in the
healthcare sector. The studies reviewed offer a solid research foundation and insightful
information on the creation of efficient predictive maintenance algorithms, integration with
current systems, privacy and security issues, and proactive maintenance techniques.
3. Proposed method
The suggested approach in this research paper concentrates on utilising machine learning
methods for proactive maintenance in the IoT for healthcare. The goal is to create precise

Predictive Maintenance in Healthcare IoT: A Machine Learning-based Approach
Section A-Research paper
ISSN 2063-5346
15912
Eur. Chem. Bull. 2023, 12(Special Issue 4), 15909-15922
models based on real-time sensor data that can forecast equipment failures or degradation in
medical devices. The suggested approach is described in the steps below:
Information Gathering: The first step entails gathering sensor information from IoT-enabled
medical devices. These devices continuously record a number of parameters, including
operating conditions, vibration, temperature, and pressure. The gathered information offers a
thorough analysis of the device's performance and can be used to inform proactive
maintenance. Data preprocessing: Prior to the development of a model, noise, missing
values, and outliers that may be present in the sensor data must be removed. To guarantee the
quality and relevance of the data, data preprocessing techniques like data cleaning,
normalisation, and feature extraction are used. Feature Selection: To create predictive
models, pertinent features from the preprocessed data are chosen in this step. In order to
determine which features are most useful for predicting equipment failure, feature selection
techniques like correlation analysis, information gain, or dimensionality reduction algorithms
are used. Model Development: Based on the chosen features, various machine learning
algorithms are applied to create predictive models. To discover the patterns and connections
between sensor data and maintenance events, supervised learning algorithms can be used,
such as decision trees, random forests, or support vector machines. Using historical data that
includes examples of both typical device operation and failure occurrences, the models are
trained. Model Evaluation: Using appropriate performance metrics like accuracy, precision,
recall, and F1-score, the developed predictive models are assessed. The evaluation measures
how well the models can forecast maintenance requirements and spot potential equipment
failures. To test the effectiveness of the models on various datasets, cross-validation
techniques may be used.
Figure.1: Figure shows the flowchart of the proposed model.

Predictive Maintenance in Healthcare IoT: A Machine Learning-based Approach
Section A-Research paper
ISSN 2063-5346
15913
Eur. Chem. Bull. 2023, 12(Special Issue 4), 15909-15922
Real-time monitoring and deployment: The predictive models are put into real-time
healthcare IoT environments after they have been trained and validated. The models provide
predictions on the likelihood of upcoming maintenance events while continuously monitoring
incoming sensor data from medical devices. The proper notifications or alerts are generated
when an anomaly or potential failure is found to start preventative maintenance procedures.
Improvement of the model over time: To adjust to shifting device conditions and increase
their accuracy over time, the predictive models should be continuously enhanced. The models
can be retrained using the most recent data and updated on a regular basis. To detect model
drift or performance degradation and initiate necessary updates or retraining, feedback
mechanisms and model monitoring techniques can also be used. The suggested approach
combines real-time sensor data with the strength of machine learning algorithms to enable
proactive and effective maintenance in healthcare IoT. Healthcare providers can reduce
downtime, improve resource allocation, and guarantee continuous patient care by anticipating
equipment failures. The suggested approach also stresses the significance of ongoing
advancement and adaptation of predictive models to keep them useful in changing healthcare
environments. The performance and efficacy of the suggested method in predictive
maintenance for healthcare IoT will be demonstrated in the following of the research paper
through experimental results and analysis.
By providing strong tools for intelligent data analysis, machine learning enables the creation
of models that can precisely diagnose particular conditions. Given that learning is a
fundamental component of intelligence, a model must be able to acquire knowledge in a way
that is comparable to that of humans. Algorithms can create classifiers that help doctors
identify patients' problems by receiving well-classified diagnostic cases. In the clinical field,
machine learning has shown to be successful, particularly when using clinical datasets to
identify and analyse diseases. Diabetes has gotten a lot of attention in clinical research using
AI because it is a common disease with a big social impact. Numerous studies have been
done in the field to identify the risk factors for diabetes, including age, blood pressure,
insulin, body mass index, and skin thickness. Patients' diabetes status has been predicted
using a variety of machine learning algorithms, including K-Nearest Neighbours (KNN),
Random Forest (RF), Artificial Neural Networks (ANN), and Decision Trees. KNN and
Logistic Regression (LR) have, among these, demonstrated superior performance in some
studies. The steps involved in predicting diabetes using machine learning and deep learning
techniques are shown in a flowchart (Figure 1) in the proposed model. For their experiments,
researchers have used platforms with GPU support, like Google Colab. Additionally,
researchers frequently produce their own datasets that are customised for the particular issue
at hand. Researchers are making progress in understanding and predicting diabetes by
utilising machine learning and creating new methodologies. The accessibility of strong tools
and platforms gives them the freedom to experiment with novel ideas and enhance the
precision of their models.

Predictive Maintenance in Healthcare IoT: A Machine Learning-based Approach
Section A-Research paper
ISSN 2063-5346
15914
Eur. Chem. Bull. 2023, 12(Special Issue 4), 15909-15922
4. Data Pre-processing
It is crucial to clean a dataset by eliminating pointless records and attributes before using it.
To find and delete such data, a methodical approach should be used. The clinic number,
episode date, and depiction are some of the attributes that must be removed from the provided
dataset in order to protect privacy. The dataset also includes missing data for some patients'
diabetes type features. It is necessary to deal with this issue because this information is
essential for the investigation focusing on the complexity of diabetes in diabetics. 26
instances in this study had missing diabetes type values, which led to their removal from the
dataset. Additionally, determining the number of missing values for each record or patient is
a crucial step in the study. It was found through testing various classifiers and missing value
percentages that removing records with more than 60% of missing values improved
performance in comparison to other methods where this problem was ignored. Researchers
can guarantee better performance and precise results in their study of diabetes by
methodically cleaning the dataset, removing irrelevant attributes, filling in the gaps left by
missing data, and optimising the dataset for analysis.
John Tukey made exploratory data analysis (EDA) popular, with the intention of motivating
analysts to delve deeper into data and produce hypotheses that could result in additional data
collection and testing. Examining model fitting and hypothesis testing assumptions, dealing
with missing values, and adjusting variables are all part of the EDA process. EDA's main
objective is to gain a thorough understanding of the data and spot any potential problems or
anomalies. It offers a methodical, scientific approach to comprehending the nature of the data
and its analysis-related implications. Importing the necessary modules and loading the
dataset are prerequisites for performing EDA in Python. The various attributes present in the
dataset are displayed in Figure 2 as an illustration of data selection in action. This process is
essential for gaining preliminary understanding of the data structure and locating potential
variables that might be of interest for additional analysis. By engaging in EDA, analysts can
better understand the data, spot patterns, outliers, and relationships, and decide on data
preprocessing, modelling strategies, and hypothesis testing with knowledge. EDA supports
the creation of valuable insights from the data and serves as a foundation for thorough data
analysis.
Figure.2: Figure shows an instance of the selected dataset.

Predictive Maintenance in Healthcare IoT: A Machine Learning-based Approach
Section A-Research paper
ISSN 2063-5346
15915
Eur. Chem. Bull. 2023, 12(Special Issue 4), 15909-15922
Implicit type conversion in Python describes when a data type is automatically changed
during execution. Type coercion is another name for this. Contrarily, explicit type
conversion, also referred to as type casting, requires the programmer to explicitly change a
variable's data type to the desired data type. Use pandas, a well-liked data manipulation
library in Python, to filter a data frame for particular columns and change the lowercase
values with uppercase in the chosen input dataset. Here is an illustration of some code:
import pandas as pd
# Load the input dataset into a data frame
df = pd.read_csv('input_dataset.csv')
# Filter specific columns
selected_columns = ['Family History', 'Age', 'Gender', 'Polyuria', 'Sudden weight loss',
'Depression stress', 'Weakness', 'Itching', 'Alopecia', 'Obesity']
filtered_df = df[selected_columns]
# Replace lowercase values with uppercase in the filtered data frame
filtered_df = filtered_df.applymap(lambda x: x.upper() if isinstance(x, str) else x)
# Generate a heat map among the data frame attributes
heatmap = filtered_df.corr()
The correlation matrix, which measures the relationships between the attributes, can be
calculated using the corr() function and used to create a heat map between the attributes of
the data frame. Using libraries like seaborn or matplotlib, the resulting correlation matrix can
be seen as a heat map. You can make a histogram plot based on the data frame's 'Age'
column to check the proportion of patients in various age groups. Here is an illustration of
some code:
import matplotlib.pyplot as plt
# Plot a histogram of the 'Age' column
plt.hist(df['Age'], bins=10)
plt.xlabel('Age')
plt.ylabel('Number of Patients')
plt.title('Distribution of Patients by Age')
plt.show()
This snippet of code plots the 'Age' column's histogram using the hist() function from the
matplotlib library. To show how many patients are in each age group, the histogram is
divided into 10 bins.

Predictive Maintenance in Healthcare IoT: A Machine Learning-based Approach
Section A-Research paper
ISSN 2063-5346
15916
Eur. Chem. Bull. 2023, 12(Special Issue 4), 15909-15922
Figure.3: Figure shows the heat map of different attributes of data and the patient's range.
Figure.4: Figure shows the histogram for different age group patients.
5. Model Training and Testing
In machine learning, a model is typically trained using a training set and its performance is
assessed using a test set. Here is a quick definition of the terms used: Data Education: The
training set in supervised learning (SL) problems consists of observations with input variables
(features) that have been recorded and an associated output variable (target). The algorithm
picks up patterns and connections between the input variables and the target variable as a

Predictive Maintenance in Healthcare IoT: A Machine Learning-based Approach
Section A-Research paper
ISSN 2063-5346
15917
Eur. Chem. Bull. 2023, 12(Special Issue 4), 15909-15922
result of these observations. To reduce prediction errors, the model's parameters or weights
are adjusted during the training process using the training data.
Figure.5: Figure shows the algorithms used for model training and testing
Data analysis: A different dataset called the test set is used to evaluate how well the trained
model performed. Any examples or observations from the training phase shouldn't be
included. This makes sure that the model is assessed using hypothetical data, giving a more
accurate assessment of its generalizability. The model's performance metrics, such as
accuracy, precision, recall, etc., can be assessed by comparing its predictions to the test set.
Six supervised learning algorithms are mentioned, which means that both the training and
testing phases used these algorithms. It's impossible to provide specific information about the
algorithms or their performance without more information or context, though. It's possible
that Figure 5 (which isn't visible in the text) offers a visual representation or comparison of
these algorithms, showcasing their features, performance metrics, or other pertinent data.
Figure.6: Random Forest classifier

Predictive Maintenance in Healthcare IoT: A Machine Learning-based Approach
Section A-Research paper
ISSN 2063-5346
15918
Eur. Chem. Bull. 2023, 12(Special Issue 4), 15909-15922
An ensemble learning method called Random Forest (RF) is frequently employed for
classification and regression tasks. During the training phase, multiple decision trees are
built, and predictions are then made based on the classes or means of the individual trees. By
randomly arranging the input features or components, RF avoids the problem of overfitting.
The pseudo-code outlines two phases that make up the RF algorithm. It is renowned for
having a high degree of precision when handling large amounts of data. RF is a supervised
algorithm that grows a number of decision trees during training and is used for classification
and regression tasks. The category that receives the most votes from the trees determines the
final prediction. By determining the best trade-off, Random Forests find a balance between
high variance and high bias. They also offer an indicator of error rates, enabling the
assessment of model effectiveness. RF algorithms are more resilient and less impacted by
such anomalies or extreme values than some AI models, such as logistic and linear
regression, which are sensitive to outliers in the training data. Due to its resilience, RF can
handle data tampering or outliers without significantly affecting the performance or accuracy
of the model.
6. Results and Discussion
A well-liked algorithm for classification tasks in data analysis and pattern recognition is
called Support Vector Machines (SVM). Finding an ideal hyperplane that successfully
divides the data points of various classes is the goal of SVM. Finding the decision boundary
that maximises the margin between classes while reducing classification error is the goal.
Support vectors, or the data points that are closest to the decision boundary, are how SVM
sets itself apart. Making precise predictions and defining the ideal hyperplane depend heavily
on these support vectors. Through the use of various kernel functions, SVM performs well
with datasets that can be separated into linear and non-linear categories. SVM divides the
entire dataset into two separate classes when two classes are being classified. There are
labelled images that are regarded as meaningful in each class and unlabeled images that are
regarded as irrelevant or noise. SVM aims to distinguish the relevant images from the
irrelevant ones and correctly classify the unlabeled images. Finding the essential and non-
essential vectors and creating a decision boundary based on them are the steps in this process.
Grid search cross-validation is a technique used by the SVM classifier to identify the ideal set
of hyperparameters. This enhances the model's functionality and capacity for generalisation.
After determining the ideal hyperparameters, the model should provide probability estimates
and be refit on the entire dataset, as indicated by the use of probability and refit with 'True'.
Figures.7 and.8 show, respectively, the analysis and predicted outcome matrices as well as
the analysis graph for the SVM classifier with predicted outcomes. These visualisations aid in
assessing the SVM classifier's effectiveness and comprehending its predictions.

Predictive Maintenance in Healthcare IoT: A Machine Learning-based Approach
Section A-Research paper
ISSN 2063-5346
15919
Eur. Chem. Bull. 2023, 12(Special Issue 4), 15909-15922
Figure.7: Predicted outcomes of SVM.
Figure.8: Frequency analysis graph for original and predicted outcomes of SVM classifier.
Figure.9: KNN predicted outcomes.

Predictive Maintenance in Healthcare IoT: A Machine Learning-based Approach
Section A-Research paper
ISSN 2063-5346
15920
Eur. Chem. Bull. 2023, 12(Special Issue 4), 15909-15922
Figure.10: Frequency analysis graph for original and predicted outcomes of KNN classifier.
SVM and KNN classifiers both achieved a specificity of 1.0 in their tests, correctly
classifying all of the negative instances, according to the results. The sensitivity of 0.0,
however, shows that they struggled to identify the positive instances with accuracy. The
accuracy of 42.95% suggests that both classifiers' overall performance is below averageIn
order to learn and extract complex patterns from data, deep learning, a subset of machine
learning, uses artificial neural networks with multiple layers. In a variety of fields, including
healthcare and medical diagnosis, deep learning has produced promising results. Deep
learning can be used to enhance the system's classifier performance and possibly achieve
higher rates of sensitivity, specificity, and accuracy. Deep learning models have the capacity
to handle more intricate relationships and identify intricate patterns in the data, which may
result in more precise predictions and diagnoses. It is crucial to remember that the
effectiveness of deep learning models depends significantly on the quantity and quality of the
data, as well as the appropriate planning and fine-tuning of the neural network architecture.
Therefore, when applying deep learning in the following section of the work, careful thought
should be given to these factors.
7. Conclusion
Machine learning techniques that enable predictive maintenance in the healthcare IoT hold
great promise for bettering maintenance procedures and enhancing patient care. This research
paper examined the use of machine learning-based predictive maintenance in the Internet of
Things (IoT) for healthcare and provided a thorough analysis of its methodology. The
classification of diabetic retinopathy using machine learning methods, particularly K-Nearest
Neighbours (KNN) and Support Vector Machine (SVM), is the main focus of the research
methodology described in the paper. The goal is to divide retinopathy images into five
categories. The proposed system performs better than conventional methods in terms of

Predictive Maintenance in Healthcare IoT: A Machine Learning-based Approach
Section A-Research paper
ISSN 2063-5346
15921
Eur. Chem. Bull. 2023, 12(Special Issue 4), 15909-15922
classification accuracy and computational efficiency, according to the study's findings. In
comparison to conventional methods, the use of machine learning techniques like KNN and
SVM allows for a more accurate and effective classification of images of diabetic
retinopathy. The proposed model can distinguish between various stages or severity levels of
diabetic retinopathy, as evidenced by the increased classification accuracy. This can be
helpful in clinical settings where timely diagnosis and management of the disease can be
aided by early detection and accurate classification of retinopathy. Additionally, the shorter
classification time indicates that the suggested system can process retinopathy images more
quickly, which is important for real-time applications or scenarios where there are many
images to analyse. Overall, the study's findings show that the proposed machine learning-
based approach outperforms conventional approaches in terms of classification accuracy and
computational efficiency, opening up a promising new path for the diagnosis and treatment of
diabetic retinopathy.
References
1. Dr. E. Baraneetharan, ''Role of Machine Learning Algorithms Intrusion Detection in WSNs: A
Survey,''J. Inf. Technol. Digit. World, vol. 02, no. 03, pp. 161–173, 2020, doi:
10.36548/jitdw.2020.3.004.
2. M. Ayad, H. Kanaan, and M. Ayache, ''Diabetes disease prediction using artificial intelligence,''Proc. -
2020 21st Int. Arab Conf. Inf. Technol. ACIT 2020, 2020, doi: 10.1109/ACIT50332.2020.9300066.
3. Nomura, M. Noguchi, M. Kometani, K. Furukawa, and T.Yoneda, ''Artificial Intelligence in Current
Diabetes Management and Prediction,''Curr. Diab. Rep.,vol.21,no.12, pp.1-6, 2021, doi:
10.1007/s11892-021-01423-2.
4. Nomura, M. Noguchi, M. Kometani, K. Furukawa, and T. Yoneda, ''Artificial Intelligence in Current
Diabetes Management and Prediction,''Curr. Diab. Rep., vol. 21, no. 12, pp. 1–6, 2021, doi:
10.1007/s11892-021-01423-2.
5. S. Shafi and G. A. Ansari, ''Early Prediction of Diabetes Disease & Classification of Algorithms Using
Machine Learning Approach,''SSRN Electron. J., no. January, 2021, doi: 10.2139/ssrn.3852590.
6. R. K. Kavitha, ''A Study on the Effectiveness of Machine Learning Algorithms in Early Prediction of
Diabetics among Patients,''Biosci. Biotechnol. Res. Commun., vol. 13, no. 11, pp. 99–104, 2020, doi:
10.21786/bbrc/13.11/22.
7. Q. Zou, K. Qu, Y. Luo, D. Yin, Y. Ju, and H. Tang, ''Predicting Diabetes Mellitus With Machine
Learning Techniques,''Front. Genet., vol. 9, no. November, pp. 1–10, 2018, doi:
10.3389/fgene.2018.00515
8. L. Kopitar, P. Kocbek, L. Cilar, A. Sheikh, and G. Stiglic, ''Early detection of type 2 diabetes mellitus
using machine learning-based prediction models,''Sci. Rep., vol. 10, no. 1, pp. 1–12, 2020, doi:
10.1038/s41598-020-68771-z.
9. J. T C, ''A Study on Various Machine Learning Classification Algorithms for Diabetes Prediction,''Int.
J. Eng. Res. Technol., vol. 10, no. 8, pp. 425–427, 2021, [Online]. Available: https://www.ijert.org/a-
study-on-various-machine-learning-classification-algorithms-fordiabetes-prediction.
10. Viloria, Y. Herazo-Beltran, D. Cabrera, and O. B. Pineda, ''Diabetes Diagnostic Prediction Using
Vector Support Machines,''Procedia Comput. Sci., vol. 170, pp. 376–381, 2020, doi:
10.1016/j.procs.2020.03.065.

Predictive Maintenance in Healthcare IoT: A Machine Learning-based Approach
Section A-Research paper
ISSN 2063-5346
15922
Eur. Chem. Bull. 2023, 12(Special Issue 4), 15909-15922
11. O. O. Oladimeji, A. Oladimeji, and O. Oladimeji, ''Classification models for likelihood prediction of
diabetes at early stage using feature selection,''Appl. Comput. Informatics, vol. ahead-of-p, no. ahead-
of-print, 2021, doi: 10.1108/aci-01-2021-0022
12. U. M. Butt, S. Letchmunan, M. Ali, F. H. Hassan, A. Baqir, and H. H. R. Sherazi, ''Machine Learning
Based Diabetes Classification and Prediction for Healthcare Applications,''J. Healthc. Eng., vol. 2021,
2021, doi: 10.1155/2021/9930985.
13. G. D. Kalyankar, S. R. Poojara, and N. V. Dharwadkar, ''Predictive analysis of diabetic patient data
using machine learning and Hadoop,''Proc. Int. Conf. IoT Soc. Mobile, Anal. Cloud, I-SMAC 2017, no.
Dm, pp. 619–624, 2017, doi: 10.1109/I-SMAC.2017.8058253.
14. N. Sneha and T. Gangil, ''Analysis of diabetes mellitus for early prediction using optimal features
selection,''J. Big Data, vol. 6, no. 1, 2019, doi: 10.1186/s40537-019-0175-6.
15. M. F. Faruque, Asaduzzaman, and I. H. Sarker, ''Performance Analysis of Machine Learning
Techniques to Predict Diabetes Mellitus,''2nd Int. Conf. Electr. Comput. Commun. Eng. ECCE 2019,
pp. 1–4, 2019, doi: 10.1109/ECACE.2019.8679365.
16. M. K. Hasan, M. A. Alam, D. Das, E. Hossain, and M. Hasan, ''Diabetes prediction using ensembling
of different machine learning classifiers,''IEEE Access, vol. 8, pp. 76516–76531, 2020, doi:
10.1109/ACCESS.2020.2989857.
17. Mujumdar and V. Vaidehi, ''Diabetes Prediction using Machine Learning Algorithms,''Procedia
Comput. Sci., vol. 165, pp. 292–299, 2019, doi: 10.1016/j.procs.2020.01.047.
18. M. Rout and A. Kaur, ''Prediction of Diabetes Risk based on Machine Learning Techniques,''Proc. Int.
Conf. Intell. Eng. Manag. ICIEM 2020, pp. 246–251, 2020, doi: 10.1109/ICIEM48762.2020.9160276.
19. W. Chen, S. Chen, H. Zhang, and T. Wu, ''A hybrid prediction model for type 2 diabetes using K-
means and decision tree,''Proc. IEEE Int. Conf. Softw. Eng. Serv. Sci. ICSESS, vol. 2017-Novem, no.
61272399, pp. 386–390, 2018, doi: 10.1109/ICSESS.2017.8342938.
20. Yahyaoui, A. Jamil, J. Rasheed, and M. Yesiltepe, ''A Decision Support System for Diabetes Prediction
Using Machine Learning and Deep Learning Techniques,''1st Int. Informatics Softw. Eng. Conf. Innov.
Technol. Digit. Transform. IISEC 2019 - Proc., no. 2, pp. 1–4, 2019, doi:
10.1109/UBMYK48245.2019.8965556.
21. K. Bhatia, S. Arora, and R. Tomar, ''Diagnosis of diabetic retinopathy using machine learning
classification algorithm,''Proc. 2016 2nd Int. Conf. Next Gener. Comput. Technol. NGCT 2016, vol. 3,
no. 2, pp. 347–351, 2017, doi: 10.1109/NGCT.2016.7877439.