Project for the Machine Learning Engineering classes at the Faculty of Electronics and Information Technologies, Warsaw University of Technology.
The aim of this project was to invent and implement webservice for delivery time predicion based on historical data.
- model training
- models exporting & loading
- serving predictions by flask restful api
- A/B experiment
- prediction summary
In /notebooks subfolder there are IPython notebooks with data exploration and models comparison steps of our project
Demo script utilizing curl run_curl_demo.sh executes scenario :
- request authors information
- load knn model
- list models
- get prediction
- load tree model
- configure A/B (A as tree B as knn)
- activate A/B experiment
- get prediction
- get prediction
- disable A/B
- set tree model as current model
- get prediction history
- get prediction summary (with responses of all models)