St.

Francis Institute of Technology 1

Department of Information Technology
Mini Project – 2B Web Based on ML (ITM 601)

MediAid: AI Medical Guide

Jigar Shah: 51
Manav Limbachiya: 52
Semira Machado: 54
Shruti Patani: 55

GUIDE
Surekha Naik Mali
Asst Professor
 2
Description

► MediAid: AI Medical Guide is a Personalized Medical Recommendation

System.
► It uses machine learning to analyze user-provided symptoms and health data.
► The system predicts potential diseases based on the input data.
► It offers tailored recommendations, including medications, workout plans,
and dietary advice.
► Designed to provide actionable and personalized healthcare guidance for
users.
 3
Need

► Personalized healthcare recommendations improve patient outcomes by

tailoring medical guidance to individual needs.
► The system ensures accurate diagnoses and provides actionable health
steps like medications, workouts, and diet advice.
► It addresses data quality challenges by cleaning and normalizing data for
optimal performance.
► It can work with different datasets, like Kaggle Diagnoses Data, to improve
accuracy and predictions.
► Supports real-time updates for dynamic and timely healthcare
recommendations.
 4
Scope

► Predicts diseases based on user symptoms using advanced ML models like

RandomForest Classifier.
► Provides tailored recommendations, including medications, workouts, and
diet plans.
► Hosted on a Flask API dashboard with real-time predictions and auto-
spelling correction.
► Targets ≥75% accuracy, limited to symptom-based predictions without
clinical data integration.
 5
Literature Review
Journal Name Year Authors Title of paper Algorithm used Result and Conclusion Limitations/ Link
Challenges

Heliyon 2023 Bhawna, A Hybrid Hybrid Filtering combines Improved accuracy and personalization Data Quality: Inconsistent or https://
Anurag Recommender collaborative and content-based over standalone methods; machine incomplete data affects www.cell.com/
Singh System for approaches, leveraging user-item learning enhanced relevance. Hybrid accuracy. Scalability: heliyon/fulltext/
Baghel Personalized interaction data and item features. system effectively combines techniques, Handling large-scale data S2405-
Healthcare Services Machine Learning techniques, offering a robust solution for efficiently.Bias and Fairness: 8440(23)10052-1
Using Machine such as clustering and personalized healthcare. Addressing biases for
Learning classification, enhance equitable treatment.
Techniques. recommendation accuracy by Integration Complexity:
analyzing user preferences and Combining techniques adds
behaviors. design complexity.

Journal of 2024 Huan TriGCN: Graph TriGCN: Combines patient, TriGCN achieved 88.17% accuracy, Integrating diverse medical https://
Systems. Zhou, Sisi Convolution disease, and medicine nodes into outperforming SVM, RF, GNN, and datasets. www.proquest.co
Personalized Liao, Network Based on a tripartite graph.GCN: traditional GCN. Overfitting risk with high m/openview/
Medicine Fanying Tripartite Graph for Propagates information using Effectively connects patients, diseases, training-to-test ratios. efc2e040c9323b3
Recommendati Guo Personalized node features and graph and medicines for personalized Longer execution times for ebd41576900899
on System. Medicine structures. healthcare recommendations. advanced models 780/1?pq-
Recommendation Calibrated Label Ranking: Ranks Addressing bias, fairness, origsite=gscholar
System. medicines based on ratings. . and security concerns. &cbl=2032325
 6
Literature Review

Journal Name Year Authors Title of paper Algorithm used Result and Conclusion Limitations/ Link
Challenges

Journal of Ambient 2021 Ching-Hsien Hsu, Interactive HSICBFO and FWSMOTE (for Improved prediction accuracy High computational https://
Intelligence and Amir H. Alavi, personalized cervical cancer risk prediction) and earlier detection of power requirements. link.springer.com/
Humanized Computing Mianxiong Dong recommendatio cervical cancer. Enhanced Need for large article/10.1007/
n systems for predictive capabilities for healthcare datasets, s12652-021-
human health identifying cervical cancer which are hard to 02950-2
risk factors. acquire/process.

De Croon R, et al. 2021 Robin De Croon, Health Knowledge-based: 35% direct use, Recommendations fell into Limited to English- https://
Health Recommender Leen Van Houdt, Recommender 65% hybrid. Hybrid.Systems: 43% four categories: lifestyle, Chinese language www.jmir.org/
Systems: Systematic Nyi Nyi Htun, Systems: adoption. Content-based: Suggested nutrition, general health care, pairs; applicability 2021/6/e18035?
Review Gregor Štiglic, Systematic educational content, activities, and specificconditions.Most to other languages is spm=5aebb161.5e
Vero Vanden Review. nutrition. Collaborative Filtering: HRSs used hybrid algorithms, untested. 1b22c2.0.0.5292c9
Abeele, Katrien Used in 3 studies. Contextual with evaluations ranging from 21yrkar9
Verbert Modeling: Added context to metrics-only to randomized
recommendations. controlled trials. HRSs
effectively guide health
actions. Five design
guidelines were proposed to
advance HRS research.
 7
Literature Review

Journal Name Year Authors Title of paper Algorithm used Result and Limitations/ Link
Conclusion Challenges

International 2020 Chaitanya Krishna SVD:Latent factor method for RBM-CNN achieved Data quality and https://
Engineering Journal for Suryadevara collaborative filtering. high accuracy, interoperability. onlinelibrary.wiley.co
Research & CRBM: Extracts features from handled large Scalability. Execution m/doi/abs/10.1111/
Development data matrices. datasets, and time for advanced exsy.12519
Towards Personalized RBM-CNN: Combines RBM provided real-time models
Healthcare - An and CNN for better accuracy. recommendations. Po Bias and fairness
Intelligent Medication tential to transform Security measures
Recommendation healthcare with against attacks.
System personalized
medication
recommendations but
requires expertise in
privacy and
integration.
 8
Problem Statement

► To provide quick and accurate health advice, MediAid predicts diseases based on symptoms and
suggests medicines, workouts, and diet plans. It helps users avoid unreliable online information
by offering real-time, easy-to-understand medical guidance with features like auto-spelling
correction and a synonym dictionary.
 9
Methodology
 10
Architectural Model
 11
Implementation

Connecting to MongoDB

Fetching the latest file from MongoDB

 12
Implementation

Using NLP
We have done extraction of date and bill
amount.
 13
Implementation

Here we have used easy ocr for

Extracting the text here we have
Used only 2 languages.
 14
Implementation

This is used for translation of one language to

another.
 15
Implementation

Here we have used pipeline model for

summarizing the text
 16
Conclusion

The Personalized Medical Recommendation System delivers tailored healthcare

guidance by predicting diseases and providing actionable recommendations like
precautions, workout plans, medications, and diet advice. It integrates advanced
machine learning models, auto-spelling correction, and a synonym dictionary for
accuracy, while a Flask API-based dashboard ensures accessibility and scalability.
Despite challenges like data cleaning, the system achieves high accuracy and
demonstrates potential as a diagnostic tool. This project streamlines personalized
healthcare, improving patient outcomes and paving the way for smarter healthcare
solutions.
 17
References

1. C.-H. Hsu, A. H. Alavi, and M. Dong, "Interactive personalized recommendation systems for human
health," Journal of Ambient Intelligence and Humanized Computing , vol. 2021, pp. 1-15, 2021.
2. H. Zhou, S. Liao, and F. Guo, "Graph Convolution Network Based on Tripartite Graph for
Personalized Medicine Recommendation System," Personalized Medicine Recommendation
System , 2024.
3. R. De Croon et al., "Health Recommender Systems: Systematic Review," Journal of Medical
Internet Research , vol. 2021, pp. 1-20, 2021.
4. Bhawna and A. S. Baghel, "A Hybrid Recommender System for Personalized Healthcare Services
Using Machine Learning Techniques," Heliyon , vol. 2023, pp. 1-10, 2023.
5. C. K. Suryadevara, "Convolutional Restricted Boltzmann Machine (CRBM) and RBM-CNN based
Health Recommender System," International Engineering Journal for Research & Development ,
vol. 2020, pp. 1-12, 2020.