A Personalized Healthcare Model for Air Pollution Monitoring and

Prediction Using Machine Learning

Gade Navaneeth, Medi Praneeth, Mohammed Khasim and Y. Mohana Roopa

Department of CSE, Institute of Aeronautical Engineering, Gandimaisamma, Hyderabad, 500043, Telangana, India

Keywords: Personalized Healthcare, Air Pollution Prediction, Web Scraping, Linear Regression, AI‑driven Health

Recommendations, Air Quality Forecasting, Machine Learning.

Abstract: Air pollution, as a major contributor to respiratory, cardiovascular, and other chronic diseases, is a significant

environmental and public health problem. This paper developed a personalized healthcare model based on air

quality prediction and individual health recommendations-finally trained on machine learning skills and based

on time-series data. Unlike traditional air quality monitoring systems that measure particulate matter and other

gases, our approach is based on a Linear Regression algorithm that predicts PM2 levels and other pollutants.

5 by merging web-scraped infectious disease information with meteorological data. The model operates as a

web-based application that allows users to enter their location to receive personalized health precautions and

real-time air quality forecasts. The guidelines, which address the estimated risk of air pollution, tell consumers

how to act preventatively, and what they should do to protect themselves, like staying indoors or using air

filtration devices, they added. By combining real-time forecasts with individual participation, the system

enables proactive health management and increases public awareness. The proposed methodology reports on

the use of AI-driven insights to facilitate an evidence-based and straightforward approach for mitigating the

negative health effects of air pollution.

1 INTRODUCTION

Air pollution, a leading cause of severe respiratory

and cardiovascular ailments, is increasingly

recognized as an important global health threat

(Wang, X., Li, L., and Chen, Z., 2020). The

intensification of human activity in urban areas and

industrial areas has already started corroding the

environment, even more so particularly in recent two

decades, not limited to air pollutants such PM2. 5,

PM10, NO₂, and SO₂ have become more serious,

aggravated the baseline risk in the population and

promote the risk of disease (P. Huang, et al., 2022).

Accurate air quality prediction is essential for

mitigating these hazards, and enabling people to take

preventive action for their health (S. S. Kumar., et al.,

2024).

This work presents a personalized healthcare

platform that integrates health support with machine

learning-based air quality prediction (S. Kumari., et

al., 2025). This approach focuses on exposure at the

individual level, rather than general environmental

monitoring, by allowing users to input their location

and receive real-time predictions of pollutants (I.

Gryech., et al., 2024). It predicts pollution levels via

Linear Regression using past meteorological data and

also scrapes the web for air quality data. To enhance

usability, the model is implemented as a web-based

application that provides users with personalized

health advice based on expected air quality

(Kekulanadara., et al., 2021). These

recommendations advise individuals to take

precautionary actions such as wearing masks, staying

indoors, or minimizing outdoor activitie during times

of high pollution levels (R. Buvana., et al., 2022).

This use of AI-driven data creates the base for a more

proactive approach to air pollution reduction, by

empowering the people to take educated health

decisions based on the information (A. Mittal., et al.,

2024).

2 RELATED WORKS

Air quality is an important health global health

problem (1) and pollutants like PM2. 5, PM10, NO₂,

and SO₂ relevant to respiratory and cardiovascular

diseases. Thanks to machine learning and artificial

Navaneeth, G., Praneeth, M., Khasim, M. and Roopa, Y. M.

A Personalized Healthcare Model for Air Pollution Monitoring and Prediction Using Machine Learning.

DOI: 10.5220/0013935800004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 5, pages

575-579

ISBN: 978-989-758-777-1

575

intelligence technology, long-term air quality

monitoring and personalized health advice by

predictive models and data-driven methods have

emerged as new avenues for research (S. B. Kasetty

and S. Nagini., 2022). We divide this literature review

into three sections that relate directly to the work in

our study: machine learning in predicting air

pollution, web scraping for data collection and AI-

powered health recommendations.

2.1 Air Pollution Prediction Using

Machine Learning

Historical weather data is often invaluable for

improving the accuracy of predictions, as machine

learning techniques have shown great promise in air

pollution forecasting A personalized healthcare

model (Behal and Singh, 2020) which integrates

various machine learning techniques to predict the

level of air pollution and its impact on parameters of

health of individuals is proposed. By utilizing both

environmental sensing data and ML algorithms, their

study improved the spatial and temporal resolution

of air quality predictions with personalized health

recommendations. The ultimate goal of the machine

learning research task is to create more accurate

predictions of air pollution and develop AI-based

health systems.

2.2 Web Scraping for Environmental

Data Collection

Web scraping is behind accurate air quality

prediction that relies on extensive datasets. Once

published these methods let alone the huge-scale

data they demand for the accurate time forecasting. It

provided an overview of the significance of web

scraping in automating data extraction processes and

converting unstructured data available on the web

into meaningful, stored, and analysable data

(Sirisuriya 2023). Their research showed how web

scraping can facilitate real-time updates to

environmental data, which helps to retrain machine

learning models using the most up-to-date pollution

metrics. Such an approach can greatly enhance the

responsiveness and accuracy of AI-based air quality

monitoring solutions.

2.3 AI-Driven Health

Recommendations Based on

Pollution Exposure

AI-driven systems are increasingly being used to

assess environmental risks and provide personalized

health interventions. Olawade et al. (2024)

highlighted how AI technologies enhance

environmental monitoring by enabling pollution

source detection, disaster forecasting, and air quality

monitoring. Their research emphasized the role of

predictive analytics in mitigating health risks,

allowing individuals to take precautionary actions

based on real-time pollution levels. Despite

challenges such as data accessibility and privacy

concerns, their study underscored the potential of AI

to revolutionize public health strategies by integrating

pollution forecasting with personalized healthcare

recommendations (Y Mohana Roopa., et al., 2023,

2022).

3 METHODOLOGY

Thus, the personalized healthcare model developed

for air pollution monitoring and prediction consists of

a methodological framework entailing data

collection, preprocessing, AQI prediction through

machine learning approaches, correlation analysis

between AQI levels and health conditions, and

finally, health recommendation generation. This

approach enables accurate prediction of air quality

while offering personalized healthcare

recommendations relative to the air pollution levels.

3.1 System Architecture

Figure 1: System Architecture.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

576

This approach leverages the latest technologies to

create an automated and intelligent air quality

monitoring system that encompasses data collection,

analysis, and predictive modelling. It includes

different modules like data acquisition, pre-

processing, machine learning model training, user

interface integration, etc. The overall system

architecture is shown in Figure 1.

3.2 Data Collection and Preprocessing

Historical and real-time air pollution data were

collected from online sources such as Open Weather

API using web scraping techniques. The dataset

includes various meteorological and pollution-related

parameters such as PM2.5, NO2, CO, temperature,

humidity, precipitation, and wind speed.

The raw dataset undergoes preprocessing,

including data cleaning, handling missing values, and

normalization. The data is then aggregated into daily

average values for the years 2013 to 2017. The

cleaned dataset is structured to ensure consistency

and readiness for machine learning analysis. Table 1

shows the Dataset Attributes Used for AQI

Prediction.

Table 1: Dataset Attributes Used for AQI Prediction.

Feature Description Unit

PM2.5

Fine particulate matter

concentration

µg/m³

NO2

Nitrogen dioxide

concentration

ppb

Carbon monoxide

concentration

ppm

Temperature

Average daily

temperature

°C

Humidity Relative humidity %

Wind speed

Wind speed at ground

level

m/s

Precipitation

Rain or snow

precipitation

3.3 Machine Learning Model for AQI

Prediction

The system employs supervised learning techniques,

specifically Linear Regression, to predict future AQI

levels. The dataset is divided into training (70%) and

testing (30%) sets, ensuring model generalizability.

Feature selection techniques are applied to retain the

most significant parameters influencing air pollution

trends.

The trained model evaluates its performance using

error metrics such as Root Mean Squared Error

(RMSE) and R² Score, ensuring the reliability of AQI

predictions. The results are visualized to compare

actual vs. predicted AQI values over time.

3.4 Personalized Healthcare

Recommendations

Using predicted AQI readings, it generates

customized healthcare recommendations. It then

offers users precautionary advice depending on how

severe the level of pollution exposure is. When

pollution levels peak, users are notified as well as

receive recommendations to reduce outdoor

activities, wear protective masks and turn on air

filtration systems.

3.5 Web-Based Deployment and User

Interaction

I mean integrated model into the platform defined in

web user models inputs the location and gets real time

model prediction with health sector prediction and

real time prediction. It securely authenticates the

users, enabling them to access environmental

conditions, pollution advisories, and health

information. This dynamic updating helps it get a

sense of how pollutants and dust flow around, so that

it can be aware of pollution in real time.

3.6 Implementation

The personalized healthcare model for air pollution

monitoring and prediction consists of three main

parts, which are data collection, ML model

development, and web deployment. The system

ingests real-time pollution data, forecasts air quality

levels, and suggests health care recommendations

It is built with Python for the data handling and

machine learning aspect of running the text analysis,

and leverages Scikit-learn for building and testing the

models. Beautiful Soup is used to scrape the web for

pollution and Meteorology data from datasets and

open Weather API. The data from pollution will be

stored in the structured form by using the SQL which

is efficient to retrieve and management of the

database. For the web-based implementation, I use

the Flask framework for the backend API and React.

as for the front-end allows the user to interact with

the system without any disturbance.

It consists of a well-defined workflow that

processes environmental data and generates custom

insights. Data Collection: Obtain pollution and

weather data from datasets and open Weather API.27

Data Pre-processing: Handle missing values if any,

A Personalized Healthcare Model for Air Pollution Monitoring and Prediction Using Machine Learning

577

and normalize the data for consistency. The model

training phase is executed using Linear Regression.

After training, the system generates AQI forecasts

along with healthcare recommendations, thus making

users aware of pollution risks and how to take

precautionary measures. The system receives input

from users via a web-based interface, in which they

share their location details and get updates on real-

time pollution in their area.

Using Pickle, the trained model is exported to be

used just fine with the Flask-based backend API. The

API enables real-time predictions by processing user

inputs to give AQI forecasts as well as personalized

health suggestions. React -- the frontend. js in this

project settles for easy-to-use interface for users to

insert their location and visualize pollution trends.

4 RESULTS AND EVALUATION

We developed and implemented a smart and

personalized healthcare model for air pollution

monitoring and prediction with the following

contributions: (1) reliable prediction of the PM2.5

AQI level and (2) personalized health advice. In the

evaluated time period the air pollution we have

trained a Linear Regression model on, are able to

determine the best combination of parameters with

air quality values. The red below the AQI prediction

graph (Figure 2) indicates smooth variance between

actual and prediction, demonstrating this system’s

predictive capability and therefore reliability in real-

world scenarios.

Figure 2: AQI Prediction Graph.

Moreover, correlation analysis between industrial

waste emissions and AQI levels, indicates that there

is a strong positive correlation between the two as

well, implying that industrial emissions contribute

significantly towards worsening air quality. This

relationship is depicted in the AQI vs Industrial

Waste Emissions Graph (Figure 3), which

demonstrates how industrial emissions contaminate

the air. The system also features a web-based user

interface to input their location, view AQI prediction

results, and get real-time health recommendations.

Figure 3: AQI and Industrial Waste Gas Emissions.

5 DISCUSSION

The system suggest which healthcare precautions

users can take, such as adjusting their mask use,

reducing their exposure to the outdoors or increasing

filtration to improve their indoor air quality; the

advice changes dynamically based on the level of

pollutants. This interactive dashboard provides a

user-friendly visualization of pollution trends, along

with any matching health advisories. While the model

can make accurate predictions, there are other areas

to improve in the future by setting up additional

machine learning models to increase prediction

accuracy, obtaining real-time pollutant sensor data

along with recommendations based on users' health

risks. This will augment the system's capacity to

combat pollution-induced health hazards and enhance

environmental awareness in general.

6 CONCLUSIONS

A personalized healthcare model for air pollution

monitoring and prediction using machine learning is

presented through this study. By investigating

historical and contemporary environmental data, the

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

578

system accurately forecasts AQI with health

advisories. The visually strong correlations between

real return AQI values and Linear Regression

predicted AQI values indicate a reliable predictive

performance, so we can go ahead and move on to air

pollution datasets training. Moreover, correlation

analysis validates the influence of industrial

discharge to the air, emphasizing the need for

pollution stormers.

It moves here closer to users by building a web-

based interface that lets users obtain health

recommendations based on the AQI in their area.

Depending on the pollution, the system decides real-

time pollution precaution measures according to

real-time pollution severity for users. The interactive

dashboard additionally increases the interaction of

users as the trends of pollution and their relevant

health risk have been displayed in a user-friendly

way.

While this model has a lot of positive impact in

terms of forecasting air pollution and providing

health advice, further enhancements can be made in

terms of exploring additional machine learning

techniques to improve model accuracy, integrating

live sensors data, and giving offer health advice

according to user-specific medical history.

REFERENCES

A. Mittal, S. Arora, S. Sharma and A. Garg, "Advancements

in Air Pollution Prediction and Classification Models:

Exploring Emerging Frontiers," 2024 10th International

Conference on Advanced Computing and

Communication Systems (ICACCS), Coimbatore,

India, 2024.

D. B. Olawade, O. Z. Wada, A. O. Ige, B. I. Egbewole, A.

Olojo, and B. I. Oladapo, "Artificial intelligence in

environmental monitoring: Advancements, challenges,

and future directions," Hygiene and Environmental

Health Advances, vol.12,2024

I. Gryech, C. Asaad, M. Ghogho, and A. Kobbane,

"Applications of Machine Learning & Internet of

Things for Outdoor Air Pollution Monitoring and

Prediction: A Systematic Literature Review,"

Engineering Applications of Artificial Intelligence, vol.

137, Part B, 2024

J. J. Bosco and V. Kowsalya, "A Novel Approach for Air

Pollution Prediction Using Machine Learning

Techniques," 2024 Third International Conference on

Electrical, Electronics, Information and Communicat-

ion Technologies (ICEEICT), Trichirappalli, India,

2024.

K. M. O. V. K. Kekulanadara, B. T. G. S. Kumara and B.

Kuhaneswaran, "Machine Learning Approach for

Predicting Air Quality Index," 2021 International

Conference on Decision Aid Sciences and Application

(DASA), Sakheer, Bahrain, 2021

P. Huang, K. Kim, and M. Schermer, "Ethical Issues of

Digital Twins for Personalized Health Care Service:

Preliminary Mapping Study," Journal of Medical

Internet Research, vol. 24, no. 1, 2022

R. Buvana, A. Sathya, R. Chathana, C. Jaysri and D. Jashna,

"Air Pollution Prediction with Machine Learning,"

2022 1st International Conference on Computational

Science and Technology (ICCST), CHENNAI, India,

2022.

S. B. Kasetty and S. Nagini, "A Survey Paper on an IoT-

based Machine Learning Model to Predict Air Pollution

Levels," 2022 4th International Conference on

Advances in Computing, Communication Control and

Networking (ICAC3N), Greater Noida, India, 2022.

S. D. S. Sirisuriya, "Importance of Web Scraping as a Data

Source for Machine Learning Algorithms - Review,"

2023 IEEE 17th International Conference on Industrial

and Information Systems (ICIIS), Peradeniya, Sri

Lanka, 2023.

S. S. Kumar, R. Chandra, and S. Agarwal, "Rule based

complex event processing for an air quality monitoring

system in smart city," Sustainable Cities and Society,

vol. 112, 2024

S. Kumari, A. K. Tyagi, S. Tiwari, and G. Kulshreshtha,

"Sensors and Digital Twin for Healthcare Monitoring

Frameworks," in Digital Twin, Blockchain, and Sensor

Networks in the Healthy and Mobile City, T. A.

Nguyen, Ed. Elsevier, 2025

Wang, X., Li, L., and Chen, Z., “Prediction of air quality

based on machine learning algorithms,” Environmental

Research, Vol. 188, pp. 109-118, 2020.

Y Mohana Roopa et, al,” Prediction Evaluation of Gene

Ontology Using Support Vector Machine",

International Journal on Recent and Innovation Trends

in Computing and Communication, June 2023, 11, pp.

522–526

Y. Mohana Roopa et.al,” Deep Learning-Based Hybrid

Framework Utilizing OpenCV and CNN for Automated

Brain Tumor Detection and MRI Image Classification”,

2024 4th International Conference on Sustainable

Expert Systems (ICSES) on 15-17 October 2024 at

Kaski, Nepal

Y. Mohana Roopa,et.al,” A secured IoT-based model for

human health through sensor data”, Measurement:

Sensors, Elsevier publisher, ISSN: 2665-9174 (Online),

Volume 24, December 2022Page No 50-58

A Personalized Healthcare Model for Air Pollution Monitoring and Prediction Using Machine Learning

579