FMLearn Model implementation using K Nearest Neighbor

FMLearn implementation using the K Nearest Neighbor algorithm, specifically sci-kit learn's KNeighborsClassifier.

This release also introduces additional features like:

• #12 force new model to be trained once data has been reloaded.
• #13 throw error if the data is not loaded before training.
• #14 check the shape of the input data frame.
• #15 force retrain of the model if the model is older than a set time frame.

And bug fixes like:

• #17 Heroku deployment failed: reason - empty database.
• #18 prediction of an algorithm when no model/dataset exists (cold start problem)
• #19 when to load the data and train model for the first time

This is also the first release after the publication of my paper, and hence an update to the website including the details of the publication.

Code restructure and UI for FMLearn

Complete restructure of code.

• moved code major chunks of code from app.py to api.py
• moved all code into src folder
• root directory only contains essentials stuff for deployment

Added UI for the server.

• provided links to GitHub Repos
• WIP for API descriptions
• and other important links

meta-features of data and kNN search

Features introduced in this release:

• added data model for meta-features
• base kNN search algorithm
• utils.py for the fetching and pre-processing of data for meta-learning

Model Parameters and Heroku Deployment

Features introduced in this Release:

• new data model (Params)
• updated FMLearn to store and retrieve model parameters
• moved all the data models to a separate folder.
• added config and constants file
• added files for deploying in Heroku (base URI: https://fmlearn.herokuapp.com/)

Draft Release v0.2

• updated README.md
• added files required for deployment in Heroku
• cleated WebHooks to auto-redeploy in Heroku when pushed to master

Draft Release v0.1

draft-v01

handled storing NUL char as a  string in db