/QRT2023

"Can you explain the price of electricity?" Competition hosted by Challenge Data and QRT

Primary LanguageJupyter Notebook

Can you explain the price of electricity? by QRT

https://challengedata.ens.fr/participants/challenges/97/

Directory structures

  • QRT2023/: Modules for DS/ML pipelines for data processing, analytics, model training, predictions in this competition
  • notebooks/: Jupyter Notebooks for running my whole workflow

Modelling Strategy 1

  1. Splitting dataset into 3 set, according to regions and data distributions
  2. Feature selection on each set, by Correlation or LASSO; And target transformation
  3. Linear models will use selected features, Tree models use all features
  4. Hyperparameters optimization by CV
  5. Ensembles of model predictions (individual model will first be bootstrapped)

Modelling Strategy 2

  1. AutoML, with or without feature engineering

Scores

Spearsman Correlation

LB Type Score
Public LB 0.295
Private LB 0.251