IEEE-CIS Second Technical Challenge Solution

- Energy prediction from smart meter data

Wenlong Wu, EECS, University of Missouri

Notebooks:

• eda-figure.ipynb: exploratory data analysis and generated visualization figures;
• tsne-visualization.ipynb: t-SNE dimension reduction and visualization;
• constant_prediction.ipynb: generate constant prediction using Nov and Dec data;
• preprocess-data-clean.ipynb: data preprocessing;
• feature-engineering.ipynb: feature engineering;
• modeling-lgb.ipynb: modeling using LightGBM model;
• mini_lgb_model.ipynb: modeling using 12 LightGBM models for each cluster;
• check_outliers.ipynb: detect suspicious meters;
• post-processing.ipynb: post processing;
• ensemble.ipynb: ensemble multiple submissions;
• visual_submission.ipynb: generate visualization figures for visualization;

Competition page:

https://ieee-dataport.org/competitions/ieee-cis-technical-challenge-energy-prediction-smart-meter-data

I ranked first position on the final evaluation:

https://engineering.missouri.edu/2020/12/mizzou-engineering-student-takes-first-place-in-ieee-computational-challenge/