Medicine Sales Prediction

Predicts medicine sales based on prior disease patterns using ML pipeline.

Predicts medicine sales demand using disease outbreak patterns, with an interactive Streamlit dashboard supporting Regression, Classification, Clustering, Anomaly Detection, Forecasting, Explainability, Trends & NLP.

📄 Project Report: View Report
📊 Live Dashboard: Launch Dashboard
🎥 Video Demo & Presentation: Watch Demo

---

📋 Table of Contents

• Overview
• Features
• Quick Start
• Project Structure
• Installation
• Model Performance
• Technical Details
• Disclaimer
• License
• Support / Contact

---

🎯 Overview

This project demonstrates how to build an end-to-end AI system for predicting medicine sales volume using disease outbreak information. It is designed for AIML students and beginners but includes modern AI modules like SHAP, SARIMA forecasting, Google Trends signals, and BERT sentiment analysis.

✅ What it does

• Loads two real datasets
  • disease_cases.csv → disease outbreak cases
  • medicine_sales.csv → medicine sales volume
• Merges datasets using date + location
• Generates outbreak features like:
  • month, year
  • disease_trend (case growth trend)
• Trains multiple ML models in one project:
  • Linear Regression, Random Forest
  • Logistic Regression (Risk prediction)
  • KMeans, DBSCAN (Clustering)
  • Isolation Forest (Anomaly detection)
  • SARIMA Forecasting
• Interactive dashboard using Streamlit

🎯 Target Audience: Students (Minor/Major project), beginners in ML/Healthcare analytics, applied AI learners

---

✨ Features

🔄 Dataset Handling

• Uses two CSV datasets
• Automatic merge using:
  • date
  • location
• Converts date into:
  • Month / Year features
• Auto-calculates:
  • disease_trend (growth % based on cases)

🤖 Machine Learning Suite (All-in-One Dashboard)

✅ Sales Prediction (Regression)

• Linear Regression
• Random Forest Regressor
  📌 Output: predicted sales_volume

✅ Sales Risk (Classification)

• Logistic Regression
  📌 Output: Low / Medium / High risk levels

✅ Clustering

• KMeans
• DBSCAN
  📌 Output: clusters of cities/regions based on case-sales patterns

✅ Anomaly Detection

• Isolation Forest
  📌 Detects abnormal spikes in disease cases or sales volume

✅ Forecasting

• SARIMA (SARIMAX) Time-series Forecasting
  📌 Forecast future sales_volume or cases

✅ Explainability

• SHAP feature importance visualization
  📌 Explains model predictions

✅ External Signals

• Google Trends analysis using PyTrends
  📌 Tracks keyword interest for outbreak-related terms

✅ NLP

• BERT Sentiment Analysis using Transformers
  📌 Analyzes public/user feedback sentiment

---

🚀 Quick Start

1) Clone & Setup

bash git clone https://github.com/editor-shannu/minor_AI cd medicine_sales_prediction python -m venv env

2. Activate Virtual Environment

✅ Windows:

env\Scripts\activate

✅ macOS/Linux:

source env/bin/activate

3. Install Dependencies pip install -r requirements.txt

4. Place Dataset CSV files

Put both CSV files inside:

medicine_sales_prediction/data/ ├── disease_cases.csv └── medicine_sales.csv

5. Launch Dashboard streamlit run app.py

Open browser:

http://localhost:8501

📁 Project Structure

medicine_sales_prediction/
├── configs/
│   ├── data.yaml
│   └── model.yaml
├── data/
│   ├── disease_cases.csv
│   └── medicine_sales.csv
├── models/
│   └── medicine_sales_model.pkl
├── notebooks/
│   ├── 01_eda.ipynb
│   └── 02_predictions.ipynb
├── scripts/
│   ├── preprocess.py
│   └── evaluate.py
├── src/
│   ├── __init__.py
│   ├── data.py
│   ├── model.py
│   └── train.py
├── app.py
├── requirements.txt
├── README.md
└── .gitignore

🛠 Installation Prerequisites

Python 3.9+ (recommended: 3.10)

Git

Step-by-Step Installation

Clone

git clone https://github.com/editor-shannu/minor_AI cd medicine_sales_prediction

Create virtual environment

python -m venv env

Activate ✅ Windows:

env\Scripts\activate

✅ macOS/Linux:

source env/bin/activate

Install packages

pip install -r requirements.txt

📊 Model Performance

Model performance depends on dataset quality and city-wise patterns.

Typical metrics shown inside dashboard:

✅ Regression: Train R² / Test R²

✅ Classification: Accuracy

✅ Clustering: Cluster distribution

✅ Anomaly: Count of anomalies detected

✅ Forecasting: Visual forecast trends

🔬 Technical Details Data Pipeline

Load disease data (disease_cases.csv)

Load sales data (medicine_sales.csv)

Parse and normalize date column

Merge using:

date

location

Feature engineering:

month

year

disease_trend = pct_change(cases) grouped by location

Train models using OOP pipeline

Features Used (ML Input)

cases

disease_trend

month

year

location_encoded

Target

sales_volume

🚨 Disclaimer

IMPORTANT: This project is for educational purposes only.

📚 Designed for learning ML and healthcare analytics

❌ Not recommended for real-world medical inventory decisions

📊 Predictions depend on dataset patterns

👨‍⚕️ Consult experts for real deployments

📄 License

This project is licensed under the MIT License.

🙏 Acknowledgments

Streamlit for dashboard UI

scikit-learn for ML models

statsmodels for SARIMA forecasting

SHAP for explainability

PyTrends for Google Trends analysis

HuggingFace Transformers for BERT sentiment analysis

📞 Support / Contact

👤 Author: Shanmukha Sri Saikumar Medisetty 📧 Email: medisettyshanmukh@gmail.com 🌐 GitHub: https://github.com/editor-shannu