Coder Social home page Coder Social logo

energy_forecast's Introduction

energy_forecast

task: energy forecasting

Data from https://www.kaggle.com/c/ashrae-energy-prediction.

Aim : 'meter_reading' prediction for different 'building_id' present at different 'site_id' for different 'meter' types.

Additional data present is weather data (e.g. temperature, wind speed, etc) as well as building data (e.g area, use of building, etc)

Preprocessing includes imputing missing data as well as removing outliers/anomalies that affect overall label value or trends in the case of time series forecasting. In our case, this would amount to figuring out anomalies in building data or site data or meter data or weather data that has a major effect in overall trend for our label.

Final modelling is done on a subset of the dataset and evaluated on RMSLE.

Feature engineering methods:

  • removing some features and evaluating models
  • utilizing all features
  • using lag values as well along with features
  • normalizing time features such as month, week and day to a sin/cos feature

Final task is evaluating on the holdout 12th month of the year with the first 11 months used for training and validation

Modelling:

  • LightGBM -> using cross validation and evaluated feature imporatance -> 0.57 score when using lag values
  • ARIMA -> used the package pmdarima; final score 6.26
  • Transformer Based Model -> used pytorch-forecasting ; final score -> 0.77

Result: for time series, gradient boosting trees present better results compared to other methods in terms of evaluation as well as evaluation time. This is also collaborated with research data on time series modelling. LightGBM is also much faster thatn XGBoost due to leaf-wise growth.

Additionally, while transformer or other deep learning methods might be better, they may not be efficient for small case scenarios.

The transformer model used is with base parameters; results might be better with additional training time

energy_forecast's People

Contributors

apratim-mishra avatar

Watchers

 avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.