ML_Service_Rest_Api

Overview

The goal of this project is to build ML system available with REST API. For building the ML service I used Django and Django Rest Framework.This machine learning web service system is designed to predict the number of subscribers of a online course after a specified time based on parameters such as:

• keywords in course title
• course price
• course level
• number of lectures
• content duration

I trained several ML models (with tuning hyperparameters) and placed them on a web server:

• LinearRegression
• RidgeRegression
• LassoRegression
• SVR with RBF kernel
• KNeighborsRegressor
• RandomForestRegressor
• GradientBoostingRegressor

Training Data

Data come from https://www.kaggle.com/andrewmvd/udemy-courses/. Dataset cointains 3600+ Udemy Course come from 4 categories:

• Business Finance
• Graphic Design
• Musical Instruments
• Web Development

Requirements

• python 3.7
• django 2.2.7
• djangorestframework 3.11.0
• pandas
• numpy
• nltk
• scikit-learn
• plotly

Content

• requirements.txt - file with all required packages.
• Analitics - folder contains all files related to analitics part of application like datasets, ml models, notebook with a description of prepare ml models and EDTA.
• ML_App - folder contains web app files.
• Images - folder contains screenshots and chars to create Readme.

How to run application(using browsable api)

• Download repository. git clone https://github.com/Cyki89/Rest_Api.git.
• Create new virtual enviroment. mkvirtualenv --python=/usr/bin/python3.7 virtualenv_name.
• Install dependecies. pip install -r requirements.txt.
• Go to main application folder. cd ML_App.
• Run server on your local machine. python manage.py runserver.
• Open localhost in your browser. https://127.0.0.1:8000. You should look Main Browable api view like this:
• Go to 'users_router'. http://127.0.0.1:8000/api/users/.
• Register new user. http://127.0.0.1:8000/api/users/register/ or use TestUser account with password: test1234.
• Go to login view. http://127.0.0.1:8000/api/users/login/. You can copy token to use token authentication in future steps.
• Go to 'requests_router. http://127.0.0.1:8000/api/requests/
• Create request. http://127.0.0.1:8000/api/requests/create/. You have to be logged in or pass valid token in headers to make a request. Fill request form and click POST button:
• See prediction. You should receive response in JSON Format:
• Then you can change the request parameters, make another request, see a list of all requests or filter the results you are interested in.

Future improvement

• Deploy this application on a public web server (currently all free servers offer insufficient capacity to store all necessary dependencies).
• Add User Interface.