How nitin-dhotre27/ml-student-graduate-dropout-analysis works
This repository predicts student dropout risk through a train-to-serve workflow: training loads student records, preprocesses features, tunes and fits a stacked classifier, and saves a deployable pipeline plus evaluation and feature-importance outputs. A FastAPI service loads that pipeline to return individual or batch predictions; the web client presents risk and model information. Synthetic-data generation supports a demonstration run. The graph follows documented and excerpted relationships; the frontend-to-API link is documented in the README, while internal frontend behavior was not sampled.
Main entry points: train.py Training Workflow · app.py FastAPI · inference.py Prediction Inference · config.py Schema and Paths
Components 15
Student Data
- Synthetic Data Generator
src/generate_synthetic_data.py - Training and Test Records
data
Model Training
- Feature Preprocessing
src/preprocessing.py - Training Workflow
src/train.py - Stacked Model Tuning
src/train.py - Schema and Paths
src/config.py
Prediction Service
- Web Client
frontend/index.html - FastAPI
backend/app.py - Prediction Inference
src/inference.py
Model Outputs
- Serialized Pipeline
models/dropout_stack_pipeline.joblib - Evaluation Metrics
models/metrics.json - Feature Importances
models/feature_importance.json
Other components
- Operator
- Student
- Risk Result
Connections 22
- Operator → Synthetic Data Generator: generates data
README.md - Synthetic Data Generator → Training and Test Records: writes records
src/generate_synthetic_data.py - Operator → Training Workflow: starts training
README.md - Training Workflow → Training and Test Records: loads records
src/train.py - Training Workflow → Feature Preprocessing: builds features
src/train.py - Training Workflow → Stacked Model Tuning: tunes and fits
src/train.py - Training Workflow → Schema and Paths: uses paths
src/train.py - Synthetic Data Generator → Schema and Paths: uses data paths
src/generate_synthetic_data.py - Feature Preprocessing → Schema and Paths: uses schema
src/preprocessing.py - Training Workflow → Serialized Pipeline: saves pipeline
src/train.py - Training Workflow → Evaluation Metrics: writes metrics
src/train.py - Training Workflow → Feature Importances: writes importances
src/train.py - Web Client → FastAPI: requests predictions
README.md - Student → Web Client: enters features
- FastAPI → Prediction Inference: requests predictions
backend/app.py - FastAPI → Schema and Paths: checks paths
backend/app.py - FastAPI → Evaluation Metrics: reads metrics
backend/app.py - Prediction Inference → Serialized Pipeline: loads pipeline
src/inference.py - Prediction Inference → Feature Importances: reads importances
src/inference.py - Prediction Inference → Schema and Paths: uses artifact paths
src/inference.py - Prediction Inference → Risk Result: returns risk
src/inference.py - FastAPI → Risk Result: serves prediction
backend/app.py