Personalized Medical Recommendation System

Dashboard Demo (click on below Dashboard video screen)

APP-DEMO_ScreenRecording_Roshni.mp4

Personalized Medical Recommendation System

Welcome to our Flask API-based Machine Learning model, designed as a comprehensive Recommendation and Prediction System. Today, I'm excited to present this one-screen dashboard that exemplifies our solution.

Our system employs a Linear Regression model at its core, focusing on creating an intelligent, user-friendly interface for medical diagnosis and recommendations. Users can input their symptoms into the dashboard, and the system will provide accurate predictions along with relevant recommendations.

Key Features:

• Comprehensive Recommendations:

  Beyond identifying potential diseases, the system offers a detailed disease description, a list of precautions, suggested workout plans, a tailored medication list, and a diet plan. This holistic approach ensures users not only understand their diagnosis but also receive actionable steps to enhance their health.

• Synonym Dictionary:

  A unique feature of our system is the synonym dictionary that maps user-friendly words to medical terms. This allows users to describe their symptoms in their own words while still receiving precise predictions.

• Auto-Spelling Correction:

  Integrated spelling correction mechanisms automatically fix any mistakes made during symptom entry, further improving the user experience.
  

Overview

The Personalized Medical Recommendation System leverages advanced machine learning techniques to provide tailored healthcare guidance. By integrating user-provided symptoms and health data, this system predicts potential diseases, recommends personalized medications, and suggests suitable workout routines. Implemented as a Flask API, it ensures accessibility and scalability across various devices.

• Link to Presentation:

• ML Model Notebook:

• Flask-APP-(windowsapi)Dashboard:

  

Technology Used

Data Science Libraries:

• scikit-learn: RandomForestClassifier, GradientBoostingClassifier, KNeighborsClassifier, SVC, MultinomialNB, StandardScaler, LabelEncoder, RFE
• Data Handling: pandas, numpy
• Visualization: matplotlib.pyplot, seaborn
• Text Processing: textblob

Web Development:

• Backend: flask

Utilities: pickle, warnings, os

ML Model Selection and Accuracy

Confusion Matrix

Runtime Before & after Data cleaning

Data Description

Symptoms Presence in the Data

# Most Frequent Diseases

Most Frequent Symptom_1, Symptom_2, Symptom_3, Symptom_4

Precaution Counts

ML The Model

This project utilizes diagnoses data to:

• Liner regression Model
• Create clear links between symptoms and diseases.
• Provide recommendations on diet and exercise according to the predicted disease.

Stats

Comparison of RFE Rank and RF Rank for Each Feature

Data Processing

• Data Loading: Loaded into Python using Pandas.
• Data Cleaning: Ensured optimal performance with Matplotlib and Seaborn.
• Model Training: Implemented using a Single Vector Machine model in a Flask-hosted app with input validation and synonym dictionaries.

Data Sources

• Kaggle Diagnoses Data: Dataset
• CBC Healthcare System: Article

Evaluation Standards

• Data Model Implementation

1. Initialization, training, and evaluation of the model
2. Data cleaning, normalization, and standardization
3. Model uses SQL or Spark data
4. Predictive power: ≥75% classification accuracy or 0.80 R-squared

• Data Model Optimization

1. Documented optimization and evaluation process
2. Model performance display

• GitHub Documentation

1. Repository cleanliness and .gitignore
2. Polished README presentation

• Presentation

1. Group member participation
2. Smooth content transitions and relevance
3. Audience engagement

Challenges and Solutions

• Data Cleaning

1. Optimization: Replaced inefficient loops with vectorized operations.
2. Logging: Added logging for cleaning steps like missing values and outliers.
3. Formatting: Standardized formats across datasets.

• Model Building

1. Optimization: Used Grid Search and Random Search for hyperparameter tuning.
2. Logging: Detailed logs for model training and validation.
3. Formatting: Consistent feature scaling and encoding.

• Dictionaries and Synonyms

1. Optimization: Used hash maps for faster synonym mapping.
2. Logging: Version control for dictionary updates.
3. Formatting: Standardized synonym formats.

• Auto-Correcting Functions

1. Optimization: Improved algorithm processing speed.
2. Logging: Tracked applied corrections for error analysis.
3. Formatting: Ensured alignment with standardized formats.

• Schema Diagrams and Relationships

1. Optimization: Normalized data for efficient queries.
2. Logging: Version control for schema changes.
3. Formatting: Regularized data types and clarified relationships.

Conclusion

AI/ML presents a significant opportunity in the healthcare industry. Despite some risks of inaccuracies, the model demonstrates high initial accuracy rates and potential for use as a diagnostic filtering mechanism.