🏥 AI-Powered Dental Appointment Prediction Suite

This repository integrates three AI-powered projects focused on dental appointment prediction:

1. Dental Appointment Prediction ML Server – A Flask backend providing REST APIs for predicting appointment likelihood and generating patient-style comments.
2. Multimodal Transformer + FNN Model – A regression model combining numeric, categorical, and textual features for accurate appointment outcome prediction.
3. Schedule.ai Frontend – A Vue.js application for collecting user input, interacting with the backend, and displaying predictions and AI-generated comments in a user-friendly interface.

---

1️⃣ Dental Appointment Prediction ML Server

The ML server is a Flask-based microservice designed for dental clinics to predict appointment likelihood and generate realistic patient messages. It exposes REST endpoints for prediction, comment generation, and hot-loading ML models and preprocessing artifacts. The server can handle numeric, categorical, and textual input, providing structured JSON responses with recommendations for clinic staff.

Key Features:

• Predicts appointment likelihood from structured patient data and free-text input.
• Generates patient-style dental comments for testing or workflow simulation.
• Hot-loadable models and preprocessing artifacts to minimize downtime.
• Modular Flask architecture with Blueprints, CORS, and robust error handling.

Workflow:

1. Client sends input to /api/predict or /api/generate-comment.
2. ModelManager loads required models and preprocessing pipelines if not already loaded.
3. Input is validated and preprocessed.
4. Model performs prediction or comment generation.
5. Output is validated and returned as structured JSON.
6. Optional hot-load updates model or artifacts without server restart.

This workflow ensures fast response, memory efficiency, and consistent API outputs for integration with frontend apps.

---

2️⃣ Multimodal Transformer + FNN Model

This is a multimodal regression model combining textual, numeric, and categorical features to predict dental appointment likelihood. The architecture uses a Transformer backbone for text embedding and a Feed-Forward Neural Network (FNN) to integrate all features for regression.

Key Features:

• Multimodal Input Handling – Incorporates patient demographics, appointment details, and free-text comments.
• Transformer Backbone – Pretrained Hugging Face models for high-quality text embeddings.
• Feed-Forward Neural Network – Integrates all features to generate a numeric likelihood score.
• Training Metrics – Tracks MSE, MAE, RMSE, R², and explained variance to monitor performance.
• Visualization – Generates loss curves, prediction plots, residual plots, and distribution charts.
• Artifact Management – Automatically saves tokenizers, scalers, trained models, and validation metrics for reproducibility.

Usage:

• Train models using a notebook with clear steps for preprocessing, training, validation, and artifact saving.
• Perform inference by loading the trained model, tokenizer, and scaler to predict on new patient data.
• Evaluate the model using metrics and visualization tools to understand performance and residuals.

The model supports both CPU and GPU usage, with GPU recommended for faster training.

---

3️⃣ Schedule.ai Frontend

Schedule.ai is a Vue.js frontend application that collects user input, communicates with the backend ML server, and presents predictions in a clean and interactive interface.

Key Features:

• Dynamic Input Collection – Users provide appointment details and patient comments through responsive forms.
• State Management – Vuex modules maintain inputs, predictions, and notifications across the app.
• API Communication – Sends structured data to the backend and receives predictions and generated comments.
• Prediction Display – AI responses are rendered in tables with clear formatting.
• Notification System – Bootstrap modals provide real-time feedback on success, warnings, or errors.
• Routing – Single-page architecture with Home, Prediction, and fallback 404 pages.

Workflow:

1. Users enter appointment data and optional hints/comments in HomeView.
2. Data is dispatched to Vuex for state management.
3. EventService sends the data to the backend ML server.
4. Predictions and comments are received and updated in Vuex.
5. PredictionView renders results in dynamic tables.
6. Any errors trigger notification modals for user feedback.

This frontend ensures a seamless user experience and clean integration with the backend ML server.

---

🏁 Summary

Together, these three projects form a full-stack solution for AI-powered dental appointment prediction:

• Backend ML Server – Provides reliable, production-ready REST APIs.
• Multimodal Model – Combines text, numeric, and categorical inputs for accurate predictions.
• Frontend Application – User-friendly interface for input, prediction visualization, and feedback.

The system is modular, scalable, and ready for deployment on local machines, Docker containers, or cloud environments, providing dental clinics with actionable insights and automation capabilities.