Please use this identifier to cite or link to this item: https://scholarbank.nus.edu.sg/handle/10635/211825
Title: REMOVE OUTLIERS AND USE ENSEMBLES OF GRADIENT BOOSTING TREES: LESSONS LEARNED FROM THE ASHRAE GREAT ENERGY PREDICTOR III KAGGLE COMPETITION
Authors: LIU HAO
Keywords: GEPIII competition
Machine learning
Frameworks and Models
Gradient Boosting Trees
Data processing steps
Issue Date: 4-Dec-2021
Citation: LIU HAO (2021-12-04). REMOVE OUTLIERS AND USE ENSEMBLES OF GRADIENT BOOSTING TREES: LESSONS LEARNED FROM THE ASHRAE GREAT ENERGY PREDICTOR III KAGGLE COMPETITION. ScholarBank@NUS Repository.
Abstract: The ASHRAE Great Energy Predictor III competition was held in late 2019 as one of the largest machine learning competitions ever held focused on building performance. It was hosted on the Kaggle platform and resulted in 39,402 prediction submissions from 3,614 teams. This paper outlines lessons learned from participants from teams who mostly scored up top 10% of the competition. Various insights were gained from their experience through an online survey, analysis of their submissions and notebooks, and the documentation of the winning teams. The contestants overwhelmingly used ensembles of various types of gradient boosting tree models to do well in the competition with the LightGBM package being the most important. They indicated that the pre-processing and feature engineering phases were the most important aspects of creating the best modelling approach. All of the survey respondents used Python as their primary modelling tool and it was common to use Kaggle Notebooks, JupyterLab, PyCharm and Notepad++ platforms as development environments. When examining the correlations between variables, it was discovered that the amount of packages used have an impact on the accuracy of the model. These conclusions are important to help steer the research and practical implementation of building energy meter prediction going forward.
URI: https://scholarbank.nus.edu.sg/handle/10635/211825
Appears in Collections:Bachelor's Theses

Show full item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
Liu Hao DBE.pdf1.89 MBAdobe PDF

RESTRICTED

NoneLog In

Page view(s)

22
checked on Feb 2, 2023

Download(s)

3
checked on Feb 2, 2023

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.