How to build a Statistical Model with Supervised Machine Learning, Train, Test, Publish and Consume that Model using Python

How to build a Statistical Model with Supervised Machine Learning, Train, Test, Publish and Consume that Model using Python

Watch the entire hands-on coding session and learn how to build a simple Machine Learning Model (Supervised Machine Learning) using Python.

  1. Simple application of Pandas for Data Frame and Numpy for Arrays
  2. SkLearn (SciKit) for using different Regressors (Decision Tree/Linear Regression)
  3. SkLearn for splitting source data set into Training and Testing datasets for cross validation
  4. Pickle for distribution of Model (file distribution)
  5. Off course my favorite, SQL Server 2019

Main purpose of this video is to show you an end-to-end development effort including development of model, creating training and testing datasets, simple ways to add additional features to improve Model accuracy, how to distribute the model and how to consume (as a end user or a business user) the model using Python

I used Azure Data Studio Notebook feature to demonstrate all the steps. You can choose your favorite Notebook tool like Jupyter or Colab or something else. Also I used Visual Studio Community edition to demonstrate how to develop end user/business user module in which the distributed model is consumed for predictions.

If you are looking for more info on where to start, I recommend you to read my previous article to get some high level insight.
http://kishoresblog.com/2019/12/05/what-is-involved-in-data-science-and-where-do-i-start/

You can take it further and improve this model. Few ideas!

1. Make this model a generic model to predict future stock price for any  Stock Symbol
2. Automate the process and write the prediction results to a table on SQL Server on a daily basis for reporting or integration with other apps.
3. Send email alerts if future stock price (predicted) hits a threshold defined buy your own Buy/Sell signal criteria