---
license: other
license_name: proprietary
license_link: LICENSE
language:
- en
pipeline_tag: tabular-classification
metrics:
- accuracy
tags:
- education
---

# FutureGuard Student Dropout Risk Prediction Models

## Overview

This repository contains the trained machine learning artifacts used by **FutureGuard**, a student dropout risk prediction system designed for educational institutions.
The models are used as part of a backend ML microservice and are **not intended to be used standalone** without the FutureGuard application stack.

The primary goal of these models is to estimate the probability of a student dropping out based on academic, attendance, and administrative indicators, and to support early intervention workflows.

---

## Models Included

This repository provides:

* **Tabular classification model**

  * Algorithm: Gradient-boosted decision trees (XGBoost)
  * Output: Dropout risk probability
* **Preprocessing artifacts**

  * Feature scaler used during training
  * Fixed feature ordering required for inference consistency

The models are consumed dynamically by the FutureGuard ML service at runtime.

---

## Intended Use

The models are intended to be used for:

* Academic risk analysis
* Early identification of at-risk students
* Supporting mentoring and counseling workflows
* Institutional analytics and reporting

They are designed to operate in conjunction with:

* Rule-based risk checks
* Structured metadata-driven feature mapping
* A FastAPI-based inference service

---

## Input & Output

### Input

* Structured tabular data representing student attributes
* Features must match the expected schema and preprocessing pipeline

### Output

* Dropout risk score (probability)
* Risk category (low, medium, high)
* Used downstream for explanations and recommendations

---

## Integration

These models are loaded at runtime by the **FutureGuard ML Service**, which:

* Downloads the model artifacts securely
* Applies preprocessing and scaling
* Combines ML predictions with rule-based logic
* Exposes a single `/predict` API endpoint

Direct inference from this repository is **not supported**.

---

## License

This repository and all included model artifacts are released under a **proprietary license**.

* Copyright © 2025 Manpreet Singh and Sumit Kumar
* All rights reserved
* Use permitted as-is
* Modification, redistribution, resale, or derivative works are prohibited without explicit permission

See the `LICENSE` file for full terms.

---

## Disclaimer

The models are provided **as-is**, without any warranties.
Predictions should be used as decision-support signals and not as the sole basis for academic or administrative actions.