Pharma analytics and patient-feedback analysis
Drug Feedback Analytics and Condition Classification
Interactive machine-learning application for analyzing patient-reported drug feedback and testing condition classification through a deployed Streamlit interface.
Project Snapshot
Compact project view
Domain
Pharma analytics and patient-feedback analysis
Project Type
Personal applied-ML project
Core ML Task
Condition classification
Dataset
Drug-feedback dataset
Technologies
Python, Pandas, Scikit-learn, Streamlit, Sentence Transformers, Matplotlib, Seaborn
Deployment
Streamlit
Problem Statement
Why this project exists
Built an interactive ML application to analyze patient-reported drug feedback and support condition classification.
Why It Matters
Why This Matters
This project connects my pharmaceutical-industry exposure with practical machine-learning work and a deployed interactive application.
My Contribution
What I built
- Built a Streamlit application for drug-feedback analysis and live prediction.
- Created EDA views for distributions, correlations, and feature relationships.
- Prepared numerical, categorical, and text-based features for ML workflows.
- Implemented classification, clustering, and regression workflows in one interface.
Workflow
From dataset to live prediction
Step 1
Dataset
Step 2
EDA and preprocessing
Step 3
Feature engineering
Step 4
Models and live prediction
Demo
Try the Live Demo
Open the deployed Streamlit application and test the prediction workflow directly.
- 1Open the app
- 2Review the overview
- 3Explore the EDA page
- 4Open the classification section
- 5Use the prediction form
- 6Test a prediction
Results
Current Application Results
The deployed Streamlit application currently reports the following results:
Classification
Test Accuracy: 69.14%
The classification workflow predicts medical-condition categories based on patient-reported drug-feedback features.
Clustering Exploration
Adjusted Rand Index: 1.00
The clustering workflow is included as an additional analytical exploration.
Regression Exploration
Mean Absolute Error: 0.0148
The regression workflow is included as an additional analytical exploration.
Classification is the main portfolio workflow. Clustering and regression are supporting exploratory components of the application.