Skip to main content

Stock Price Prediction

By Mrigank Pandey

An in-house project focused on the use case of stock price forecasting

OBJECTIVES

  • Develop an end-to-end solution for prediction of stock price using various forecasting algorithms.

  • Prepare an extensive technical and deployment framework that can be used as a ready reference for building similar solutions and for presenting capability demos to other potential vendors

  • Explore and implement various concepts including automation, dashboarding, deployment and monitoring

TECHNOLOGIES USED

Databricks, MLFlow, Azure DevOps

SOLUTION APPROACH

  • Scraped data of a stock for past 1 year from Money Control API

  • Data cleaning in bronze layer of Databricks

  • Stored pre-processed data in silver layer and created a feature store.

  • Using feature store, created different SOTA forecasting models such as TFT, N-beats, Deep-AR.

  • Used MLFlow for model development, experiment tracking, model management and deployment.

  • Compared multiple metrics such as RMSE, MAE, R2 Score etc. to select best performing model

  • Stored prediction in delta table and created dashboard to visualize KPIs.

Read next