Date of publication XXXX 00, 0000, date of current version XXXX 00, 0000

particularly useful in the healthcare industry, where data 
privacy and security are of paramount importance. With 
federated learning, sensitive patient data can be kept on 
individual devices and hospitals' servers, rather than being 
centralized in a single location [13,14, 15]. This can help to 
protect patient privacy and comply with regulations such as 
HIPAA [16]. In addition, federated learning can enable the 
training of more accurate models by allowing for the 
aggregation of data from a larger number of patients. This can 
be especially beneficial in rare disease research [17], where a 
centralized dataset may not have enough examples to train a 
reliable model. Federated learning can also enable the training 
of models on a more diverse patient population, which can 
lead to more generalizable and therefore more useful models 
[18]. 
In this study, we conducted clinical event classification 
using vital signs data with Federated Learning. The Flower FL 
algorithm was selected for the implementation of keeping the 
privacy of the dataset. Several machine learning techniques 
such as Random Forest regression, XGBoots, and other 
models were used to compare to get the optimal result. In 
federated learning, we used our custom fine-tuned method to 
get the best results. In section 2, Related works are 
represented, and Section 3 is Material and Methods. Before 
Section 4 Experimental results, the last section 5 is 
Conclusion. 
II.
BACKGROUND 
Clinical event classification using vital signs data is a critical 
task in healthcare as it allows for early detection and 
management of various medical conditions. Many researchers 
around the world have studied the use of computational 
techniques to predict how long patients will stay in the hospital 
[19]. 

---

Ruzaliev R: 
Federated Learning for Clinical Event Classification Using Vital 
Signs Data 
2 
VOLUME XX, 2023 
Machine learning is a popular approach in this field, as it 
allows for the analysis of vast amounts of historical and 
current data from various sources in healthcare to make 
predictions about the future [1,20]. Medical machine learning 
contributes greatly to reducing the investment spent on it and 
to renewing the relationship between doctor and patient by 
reducing investment in it [21]. A wireless radar, for example, 
collects vital sign data using radar technology and categorizes 
healthy and infected people using five machine-learning 
models [22]. In 2019 years, Juan-Jose Beunza et al [23], to 
predict clinical events, compared several supervised 
classification machine learning algorithms for internal validity 
and accuracy. The Framingham open database used new 
methods in the data preparation process and get women an 
accuracy value of 0.81 while men had a value of 0.78. 
However, their performance in the degree of accuracy is not 
considered sufficient and is often hindered by the lack of large, 
diverse, and labeled data. Yuanyuan et al [24] introduced the 
system for using a convolutional neural network (CNN) with 
enhanced deep learning techniques to predict heart disease on 
an Internet of Medical Things (IoMT) platform. The 
"enhanced deep learning" aspect likely refers to using 
advanced techniques such as trans-fer learning or ensemble 
methods to improve the performance of the CNN. The IoMT 
platform refers to the use of medical devices connected to the 
internet to collect and transmit data for analysis. 
As we proposed enhanced clinic event classification method 
with Federated learning is a distributed machine learning 
technique that addresses this issue by allowing multiple 
devices, such as wearables and medical devices, to train a 
shared model on their own data while preserving the privacy 
of the patients. For example, Jie Xu et al [12] wrote the survey 
aims to examine the use of federated learning in the 
biomedical field. It will provide an overview of the various 
solutions for dealing with statistical, system, and privacy 
challenges in federated learning. Another example is 
highlighting the potential applications and impact of these 
technologies in healthcare. Another research in this field is that 
Thanveer Shaik et al [25] work proposes a decentralized 
privacy-protected system for monitoring in-patient activity in 
hospitals using sensors and AI models to classify 12 routine 
activities with FedStack system. FedStack is a proposed 
system for using stacked federated learning for personalized 
activity monitoring. Federated learning is a technique for 
training machine learning models on decentralized data, where 
data is distributed across multiple devices or locations. 
Stacked federated learning refers to a specific technique where 
multiple federated models are trained and then combined to 
form a final model. This paper suggests using this approach 
for activity monitoring, which likely involves collecting data 
from sensors or other devices worn by individuals to track 
their physical activity and using the trained models to 
personalize the monitoring and analysis of that data. Similarly, 
Ittai Dayan at all [26] worked on predicting the future oxygen 
requirements for symptomatic COVID-19 patients using vital 
signs, laboratory data, and chest X-rays with the FL model. 
Also, the research proposes using federated learning for 
predicting clinical outcomes in patients with COVID-19. 
Federated learning is a technique for training machine learning 
models on decentralized data, where data is distributed across 
multiple devices or locations. In this case, the authors suggest 
using this approach to train models on data from different 
hospitals or clinics, in order to improve the accuracy of pre-
dictions for patients with COVID-19. They also claim that this 
approach can be useful to make predictions in real time and 
that it can be useful to improve the performance of the models 
by sharing knowledge across different institutions. 
As we combined and improved the related approaches 
mentioned above that our pro-posed method includes more 
advantages such as privacy by training models on 
decentralized data, federated learning allows for the protection 
of sensitive patient information, as data never leaves the 
individual devices or institutions or robustness as Federated 
learning allows for the integration of data from different 
sources, which can lead to more robust and accurate models.