Pune Vidyarthi Griha’s
COLLEGE OF ENGINEERING, TECHNOLOGY AND MANAGEMENT,
PUNE-09
Department of Information Technology.
B.E. IT PROJECT
[Academic Year 2025-26] Date: 03/09/25
SYNOPOSIS
PROJECT TITLE: Predictive Early Warning System for Multi-Disease Co-
occurrence in Young Adults using XAI
ABSTRACT
Healthcare prediction systems are often designed for a single disease, but in reality, young adults
are increasingly prone to multiple lifestyle-related diseases such as diabetes, heart disease, liver
disorders, and kidney problems. This project proposes a Predictive Early Warning System that can
forecast the likelihood of multiple diseases simultaneously. Using machine learning models such
as Random Forest and XGBoost, and enhanced with Explainable AI (XAI) techniques like SHAP,
the system will not only predict diseases but also provide transparency by explaining which health
factors contributed to the prediction. This tool can help young adults take preventive measures and
assist healthcare providers in decision-making.
Keywords: Multi-disease prediction, Machine Learning, Explainable AI, Random Forest,
XGBoost, Healthcare, GRAD-CAM , SHAP.
INTRODUCTION
In recent years, the prevalence of chronic health conditions among young adults has increased
significantly due to lifestyle changes, poor dietary habits, sedentary routines, and rising stress
levels. Traditionally, most healthcare prediction systems are designed to detect or predict a single
disease, such as diabetes or heart disease. However, medical research has shown that these
conditions often do not occur in isolation. Instead, they tend to co-occur and interact with each
other, for example: diabetes increases the risk of kidney and heart problems, while liver issues
may arise due to metabolic disorders.
This co-occurrence of diseases creates a challenge for both individuals and healthcare providers,
as single-disease prediction systems fail to give a holistic picture of a person’s health risks. There
is a growing need for a system that can predict multiple diseases simultaneously and offer
meaningful insights into the relationships between health parameters and disease likelihood.
Our project, “Predictive Early Warning System for Multi-Disease Co-occurrence in Young
Adults using Explainable AI (XAI),” aims to fill this gap. The system leverages machine learning
models such as Random Forest and XGBoost to predict the probability of multiple diseases like
diabetes, heart disease, liver disease, and kidney disease in young adults. Unlike traditional “black-
box” AI models, this system incorporates Explainable AI techniques (e.g., SHAP) to provide
clear justifications for each prediction.
This ensures that users and healthcare professionals not only receive accurate predictions but also
understand why those predictions were made (e.g., “high glucose and BMI significantly contribute
to diabetes risk”). By combining predictive accuracy with transparency, the system aims to serve
as a valuable early-warning tool that empowers young adults to take preventive measures and
promotes trust in AI-driven healthcare solutions.
LITERATURE SURVEY
Research Paper Seed Idea Future Scope
Explainable AI Framework Integrated SHAP/LIME to explain disease Extendable to multi-disease
for NCD Prediction (2021, prediction models, mainly for single settings, where XAI explains
IEEE Access) diseases like diabetes. co-occurrence predictions for
multiple conditions at once.
Multi-Disease Prediction Applied LSTM deep learning on hospital Extendable to lightweight
using LSTM RNNs (2021, EHRs to predict multiple future models (Random Forest,
Expert Systems with diagnoses. XGBoost) that can work with
Applications) smaller public datasets and still
predict multiple diseases
effectively.
FedHealth: A Federated Proposed federated + transfer learning Extendable to a simplified
Transfer Learning using wearable healthcare data to ensure multi-disease prediction system
Framework for Wearable privacy. without complex federated
Healthcare (2019) setup, focusing on young adults
with public datasets..
Comparative Studies on Compared ensemble models like Decision Extendable by combining
Ensemble ML for Trees, Random Forest, and XGBoost for ensemble models with XAI and
Healthcare (2019–2023, disease prediction. deploying them in a user-
IEEE Proceedings) friendly app for real-world
preventive healthcare.
PROPOSED WORK
Model Development
The predictive model will be developed using multiple machine learning algorithms such
as Random Forest. These models are chosen because they handle tabular healthcare
datasets effectively, provide high accuracy, and allow feature importance analysis. The
dataset will include health parameters such as age, gender, BMI, blood pressure, glucose,
cholesterol, hemoglobin, and serum creatinine.
The system will be trained for multi-label classification, meaning it can predict the risk
of more than one disease (diabetes, heart disease, liver disease, kidney disease)
simultaneously. Performance will be measured using metrics such as Accuracy, Precision,
Recall, F1-score, and ROC-AUC, ensuring that the model is both reliable and well-
balanced.
System Development
Backend: Developed using Flask or FastAPI, the backend will process user inputs, interact
with the trained ML model, and return predictions. Develope user configurations for
parameter definitions and weighting, a real-time dashboard, and a maintenance activity log
interface.
Frontend: The frontend will provide an interactive interface where users can enter their
health details (e.g., age, glucose level, blood pressure)..
The workflow will follow a simple pipeline: user inputs → backend ML model →
prediction + explanation → displayed on the UI.
Advanced Features
Explainable AI (XAI): Integration of SHAP (SHapley Additive exPlanations) to provide
interpretability. Users will be able to see which factors contributed most to their predicted
disease risk.
Multi-Disease Co-Occurrence Prediction: Unlike traditional models focusing on one
disease, this system can simultaneously predict multiple diseases and show possible
interrelations.
Scalability: The system is designed in a modular way, allowing additional diseases to be
included in future versions
SYSTEM REQUIREMENTS
a) Hardware:
Processor: Intel i5 or higher
RAM: 8 GB or higher
Storage: 1 TB HDD / 256 GB SSD
GPU (optional, for deep learning): NVIDIA CUDA-enabled
b) Software:
Language: Python 3.9+
Libraries: Pandas, NumPy, Scikit-learn, XGBoost, SHAP, Matplotlib
Backend: Flask / FastAPI
Frontend: [Link] or Streamlit
Tools: Jupyter Notebook, GitHub, Docker
EXPECTED RESULT
A web-based system that predicts the probability of multiple diseases simultaneously.
Explanations for each prediction (e.g., “High Glucose and BMI contributed most to
Diabetes risk”).
Improved preventive healthcare awareness among young adults.
Potential future integration with wearable devices for real-time health monitoring.
REFERENCES
1. Ahad Rehman et al., “Multiple Disease Prediction System using Machine Learning,”
JIEEE, 2024.
2. Khishigsuren Davagdorj et al., “Explainable AI based Framework for Non-
Communicable Diseases Prediction,” IEEE Access, 2021.
3. J. Ma et al., “Multi-disease prediction using LSTM recurrent neural networks,” ESWA,
2021.
4. D. Li et al., “Context-aware Health Event Prediction via Dynamic Disease Graphs,”
arXiv, 2021.
5. Comparative ML studies on multiple disease detection, IEEE Proceedings, 2019–2023.
Project Guide Project Coordinator
Prof. [Link] Prof. [Link]