Catalyst Crew SmartPHC: AI-Driven PHC Inventory Management

Catalyst Crew SmartPHC is an intelligent, facility-level platform developed by Catalyst Crew, designed to enhance operational efficiency at Primary Health Care (PHC) centers. It combines data synthesis, predictive modeling, database management, interactive dashboards, and a multilingual AI chatbot to enable data-driven decision-making and streamline inventory management.

---

Table of Contents

• Overview
• Key Features
• Technologies Used
• Data Structure
• Installation & Setup
• Usage
• Future Enhancements

---

Overview

SmartPHC addresses PHC inventory management challenges by:

1. Synthesizing realistic inventory and patient flow data (~10,000 rows, ~27 years).
2. Predicting drug stock levels 7 days in advance using an LSTM model.
3. Loading and managing data in a relational database.
4. Providing an interactive inventory form for daily data submission.
5. Displaying insights through an embedded Power BI dashboard.
6. Offering a multilingual AI chatbot fine-tuned for medical and operational queries.

---

Key Features

1. Data Synthesis

Generates PHC inventory data including:

• date
• Total_Patients
• OutPatients
• InPatients
• Medical_Staff
• Total_Beds
• Beds_Occupied
• Beds_Available
• Malaria_Drugs_In_Stock
• Typhoid_Drugs_In_Stock
• Cough_Drugs_In_Stock
• Cold_Drugs_In_Stock
• Paracetamol_Drugs_In_Stock

---

2. Time Series Forecasting (LSTM)

• Predicts drug quantities after 7 days.
• Performs feature engineering and data transformation prior to training.
• Supports proactive inventory management and prevents stockouts.

---

3. Database Management

• Stores synthesized and user-submitted data.
• Enables efficient querying for analytics and predictive modeling.

---

4. Inventory Form

• Interactive form to submit daily PHC data.
• Collects the same variables as the synthesized dataset.
• Automatically stores data in the database.

---

5. Inventory Form Submission & Prediction

• Retrieves the last 7 days of submitted data.
• Generates drug stock predictions using the trained LSTM model.

---

6. Power BI Dashboard

• Embedded dashboard visualizes inventory and operational trends.
• Supports data-driven decision-making for administrators.

---

7. Multilingual AI Chatbot

• Fine-tuned Gemini LLM for medical context and conversation history.
• Understands follow-up questions for accurate responses.
• Supports multiple languages for diverse PHC staff.

---

Technologies Used

• Backend: Python, Streamlit
• Database: MySQL
• Machine Learning: TensorFlow, Keras, LSTM
• Data Visualization: Power BI
• AI & NLP: Gemini LLM fine-tuning
• Data Processing: Pandas, NumPy
• Others: Joblib, dotenv

---

Data Structure

Feature	Type
date	Date
Total_Patients	Integer
OutPatients	Integer
InPatients	Integer
Medical_Staff	Integer
Total_Beds	Integer
Beds_Occupied	Integer
Beds_Available	Integer
Malaria_Drugs_In_Stock	Integer
Typhoid_Drugs_In_Stock	Integer
Cough_Drugs_In_Stock	Integer
Cold_Drugs_In_Stock	Integer
Paracetamol_Drugs_In_Stock	Integer

---

Installation & Setup

1. Clone the repository:

   git clone https://github.com/Ibe-Donald/Catalyst_Crew
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Configure your gemini api_key, and database connection in .env.

4. Load the csv data to your database:

   python loading_csv_to_database.py
   

5. Create your PowerBI dashboard, embed your report and use embed link in PowerBI_Dashboard.py

6. Run the Streamlit app:

   streamlit run app.py
   

Usage

• Submit Inventory Data: Enter daily PHC inventory information using the Inventory Form.
• Automatic Database Storage: All submitted data is securely stored in the connected database for tracking and analysis.
• Predictive Analytics: The LSTM model forecasts future drug stock levels to prevent stockouts and ensure smooth operations.
• Insights Dashboard: Access operational and inventory trends through the embedded Power BI Dashboard.
• Multilingual Chatbot Support: Interact with the AI chatbot for guidance, operational support, and follow-up queries in multiple languages.

---

Future Enhancements

• Expanded Predictive Models: Introduce models for patient flow, bed occupancy, and other operational metrics.
• Real-Time Notifications: Alerts for low stock levels or critical inventory shortages.
• Mobile Application: Offline-capable app for field staff to enter data on-the-go.
• Enhanced Chatbot Knowledge: Broaden the chatbot’s medical knowledge base for more comprehensive support and accurate responses.