BuildProg: Program Generation for Testing ML-based Building Load Forecasting models via LLM and Prompt Engineering

Yang Deng; Donghua Xie; RUI Liang; Jingyun Zeng; Samson Kin Hon Tai; Dan Wang

doi:10.1145/3671127.3698698

BuildProg: Program Generation for Testing ML-based Building Load Forecasting models via LLM and Prompt Engineering

Yang Deng, Donghua Xie, RUI Liang, Jingyun Zeng, Samson Kin Hon Tai, Dan Wang

Department of Accountancy, Economics and Finance

Research output: Chapter in book/report/conference proceeding › Conference proceeding › peer-review

Abstract

Machine learning-based building load forecasting (BLF) is crucial for the building automation community, and numerous ML models have been developed for this purpose. However, a significant challenge arises when promoting these models for deployment in real buildings: building practitioners often struggle with ML-related programming. To address this issue, we propose BuildProg, a program generation tool that leverages prompt engineering to decompose user requirements and guide large language models (LLMs) in generating the necessary Python code. In its current version, BuildProg supports four tasks related to the testing of BLF models.

Original language	English
Title of host publication	BuildSys '24: Proceedings of the 11th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation
Place of Publication	New York, N
Publisher	Association for Computing Machinery (ACM)
Pages	248-249
Number of pages	2
ISBN (Print)	9798400707063
DOIs	https://doi.org/10.1145/3671127.3698698
Publication status	Published - 29 Oct 2024
Event	11th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation - Hangzhou, China Duration: 7 Nov 2024 → 8 Nov 2024 https://dl.acm.org/doi/proceedings/10.1145/3671127 (Conference proceedings)

Publication series

Name	BUILDSYS BuildSys: Systems for Energy-Efficient Buildings, Cities, and Transportation
Publisher	Association for Computing Machinery

Conference

Conference	11th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation
Country/Territory	China
City	Hangzhou
Period	7/11/24 → 8/11/24
Internet address	https://dl.acm.org/doi/proceedings/10.1145/3671127 (Conference proceedings)

User-Defined Keywords

LLM
Model testing
program generation
prompting

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1145/3671127.3698698

Cite this

Deng, Y., Xie, D., Liang, RUI., Zeng, J., Tai, S. K. H., & Wang, D. (2024). BuildProg: Program Generation for Testing ML-based Building Load Forecasting models via LLM and Prompt Engineering. In BuildSys '24: Proceedings of the 11th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation (pp. 248-249). (BUILDSYS BuildSys: Systems for Energy-Efficient Buildings, Cities, and Transportation). Association for Computing Machinery (ACM). https://doi.org/10.1145/3671127.3698698

Deng, Yang ; Xie, Donghua ; Liang, RUI et al. / BuildProg: Program Generation for Testing ML-based Building Load Forecasting models via LLM and Prompt Engineering. BuildSys '24: Proceedings of the 11th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation. New York, N : Association for Computing Machinery (ACM), 2024. pp. 248-249 (BUILDSYS BuildSys: Systems for Energy-Efficient Buildings, Cities, and Transportation).

@inproceedings{25571efb46e54304909ec6febc6f5850,

title = "BuildProg: Program Generation for Testing ML-based Building Load Forecasting models via LLM and Prompt Engineering",

abstract = "Machine learning-based building load forecasting (BLF) is crucial for the building automation community, and numerous ML models have been developed for this purpose. However, a significant challenge arises when promoting these models for deployment in real buildings: building practitioners often struggle with ML-related programming. To address this issue, we propose BuildProg, a program generation tool that leverages prompt engineering to decompose user requirements and guide large language models (LLMs) in generating the necessary Python code. In its current version, BuildProg supports four tasks related to the testing of BLF models.",

keywords = "LLM, Model testing, program generation, prompting",

author = "Yang Deng and Donghua Xie and RUI Liang and Jingyun Zeng and Tai, {Samson Kin Hon} and Dan Wang",

note = "Funding Information: Dan Wang{\textquoteright}s work is supported by RGC GRF 15200321, 15201322, 15230624, RGC-CRF C5018-20G, ITC ITF-ITS/056/22MX, and PolyU 1-CDKK, G-SAC8. Publisher copyright: {\textcopyright} 2024 Copyright held by the owner/author(s).; 11th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation ; Conference date: 07-11-2024 Through 08-11-2024",

year = "2024",

month = oct,

day = "29",

doi = "10.1145/3671127.3698698",

language = "English",

isbn = "9798400707063",

series = "BUILDSYS BuildSys: Systems for Energy-Efficient Buildings, Cities, and Transportation",

publisher = "Association for Computing Machinery (ACM)",

pages = "248--249",

booktitle = "BuildSys '24: Proceedings of the 11th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation",

address = "United States",

url = "https://dl.acm.org/doi/proceedings/10.1145/3671127",

}

Deng, Y, Xie, D, Liang, RUI, Zeng, J, Tai, SKH & Wang, D 2024, BuildProg: Program Generation for Testing ML-based Building Load Forecasting models via LLM and Prompt Engineering. in BuildSys '24: Proceedings of the 11th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation. BUILDSYS BuildSys: Systems for Energy-Efficient Buildings, Cities, and Transportation, Association for Computing Machinery (ACM), New York, N, pp. 248-249, 11th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation, Hangzhou, China, 7/11/24. https://doi.org/10.1145/3671127.3698698

BuildProg: Program Generation for Testing ML-based Building Load Forecasting models via LLM and Prompt Engineering. / Deng, Yang; Xie, Donghua; Liang, RUI et al.
BuildSys '24: Proceedings of the 11th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation. New York, N: Association for Computing Machinery (ACM), 2024. p. 248-249 (BUILDSYS BuildSys: Systems for Energy-Efficient Buildings, Cities, and Transportation).

Research output: Chapter in book/report/conference proceeding › Conference proceeding › peer-review

TY - GEN

T1 - BuildProg: Program Generation for Testing ML-based Building Load Forecasting models via LLM and Prompt Engineering

AU - Deng, Yang

AU - Xie, Donghua

AU - Liang, RUI

AU - Zeng, Jingyun

AU - Tai, Samson Kin Hon

AU - Wang, Dan

N1 - Funding Information: Dan Wang’s work is supported by RGC GRF 15200321, 15201322, 15230624, RGC-CRF C5018-20G, ITC ITF-ITS/056/22MX, and PolyU 1-CDKK, G-SAC8. Publisher copyright: © 2024 Copyright held by the owner/author(s).

PY - 2024/10/29

Y1 - 2024/10/29

N2 - Machine learning-based building load forecasting (BLF) is crucial for the building automation community, and numerous ML models have been developed for this purpose. However, a significant challenge arises when promoting these models for deployment in real buildings: building practitioners often struggle with ML-related programming. To address this issue, we propose BuildProg, a program generation tool that leverages prompt engineering to decompose user requirements and guide large language models (LLMs) in generating the necessary Python code. In its current version, BuildProg supports four tasks related to the testing of BLF models.

AB - Machine learning-based building load forecasting (BLF) is crucial for the building automation community, and numerous ML models have been developed for this purpose. However, a significant challenge arises when promoting these models for deployment in real buildings: building practitioners often struggle with ML-related programming. To address this issue, we propose BuildProg, a program generation tool that leverages prompt engineering to decompose user requirements and guide large language models (LLMs) in generating the necessary Python code. In its current version, BuildProg supports four tasks related to the testing of BLF models.

KW - LLM

KW - Model testing

KW - program generation

KW - prompting

U2 - 10.1145/3671127.3698698

DO - 10.1145/3671127.3698698

M3 - Conference proceeding

SN - 9798400707063

T3 - BUILDSYS BuildSys: Systems for Energy-Efficient Buildings, Cities, and Transportation

SP - 248

EP - 249

BT - BuildSys '24: Proceedings of the 11th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation

PB - Association for Computing Machinery (ACM)

CY - New York, N

T2 - 11th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation

Y2 - 7 November 2024 through 8 November 2024

ER -

Deng Y, Xie D, Liang RUI, Zeng J, Tai SKH, Wang D. BuildProg: Program Generation for Testing ML-based Building Load Forecasting models via LLM and Prompt Engineering. In BuildSys '24: Proceedings of the 11th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation. New York, N: Association for Computing Machinery (ACM). 2024. p. 248-249. (BUILDSYS BuildSys: Systems for Energy-Efficient Buildings, Cities, and Transportation). doi: 10.1145/3671127.3698698

BuildProg: Program Generation for Testing ML-based Building Load Forecasting models via LLM and Prompt Engineering

Abstract

Publication series

Conference

User-Defined Keywords

UN SDGs

Access to Document

Fingerprint

Cite this