This project provides a REST API built with Flask for running machine learning predictions to identify whether a patient presents an orthopedic anomaly based on six biomechanical attributes.
The API supports single and batch predictions, lists available models, and allows selecting which model is active for inference.
A key feature of this service is that ML models are stored and retrieved using MongoDB GridFS, enabling efficient management of multiple models and versions.
Two models are currently available:
- XGBoost classifier
- Linear Regression classifier
The full training pipelines, preprocessing steps, evaluation, and dataset exploration can be found in a separate project:
🔗 https://github.com/SebastianGranadosJ/Orthopedic-Anomaly-Detection-MlModel
- Predict orthopedic anomalies from biomechanical data
- Single prediction and batch prediction support
- Models stored in MongoDB GridFS
- List all available models
- Activate specific models
- Works with both XGBoost and scikit-learn models
Below is an overview of the main API endpoints used to interact with machine learning models stored in MongoDB GridFS and perform predictions using the currently active model.
Retrieves a list of all stored models.
Returns basic information such as the file ID, filename, model name, version, and upload date.
Useful for identifying which models are available in the system.
Activates a specific model so it becomes the one used for inference.
Requires sending model_name and version in the JSON body.
Automatically deactivates any previously active model and marks the selected one as the new active model.
Runs a single prediction using the currently active model.
Requires sending a features array in the JSON body, containing the six biomechanical attributes expected by the model.
Returns the model name, version, and the resulting prediction probability or value.
Executes batch predictions using the active model.
Requires uploading a CSV file containing multiple rows of the six biomechanical attributes.
The API processes all rows, appends the predictions, and returns a downloadable CSV file with the results.
📂 UCI Machine Learning Repository
(Dua, D. & Graff, C., 2019)