An AI Framework for Modelling and Evaluating Attribution Methods in Enhanced Machine Learning Interpretability

Alfredo Cuzzocrea; Qudrat E. Alahy Ratul; Islam Belmerabet; Edoardo Serra

doi:10.1109/COMPSAC57700.2023.00158

An AI Framework for Modelling and Evaluating Attribution Methods in Enhanced Machine Learning Interpretability

Alfredo Cuzzocrea, Qudrat E. Alahy Ratul, Islam Belmerabet, Edoardo Serra

Computer Science Department

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

Abstract

In this paper, we propose a general methodology for estimating the degree of the attribution methods precision and generality in machine learning interpretability. Additionally, we propose a technique to measure the attribution consistency between two attribution methods. In our experiments, we focus on the two well-known model agnostic attribution methods, SHAP and LIME, then we evaluate them on two real applications in the attack detection field. Our proposed methodology highlights the fact that both LIME and SHAP are lacking precision, generality, and consistency. Therefore, more inspection is needed in the attribution research field.

Original language	English
Title of host publication	Proceedings - 2023 IEEE 47th Annual Computers, Software, and Applications Conference, COMPSAC 2023
Editors	Hossain Shahriar, Yuuichi Teranishi, Alfredo Cuzzocrea, Moushumi Sharmin, Dave Towey, AKM Jahangir Alam Majumder, Hiroki Kashiwazaki, Ji-Jiang Yang, Michiharu Takemoto, Nazmus Sakib, Ryohei Banno, Sheikh Iqbal Ahamed
Publisher	IEEE Computer Society
Pages	1030-1036
Number of pages	7
ISBN (Electronic)	9798350326970
DOIs	https://doi.org/10.1109/COMPSAC57700.2023.00158
State	Published - 2023
Event	47th IEEE Annual Computers, Software, and Applications Conference, COMPSAC 2023 - Hybrid, Torino, Italy Duration: 26 Jun 2023 → 30 Jun 2023

Publication series

Name	Proceedings - International Computer Software and Applications Conference
Volume	2023-June
ISSN (Print)	0730-3157

Conference

Conference	47th IEEE Annual Computers, Software, and Applications Conference, COMPSAC 2023
Country/Territory	Italy
City	Hybrid, Torino
Period	26/06/23 → 30/06/23

Keywords

Artificial Intelligence
Intelligent AI Tools
Machine Learning Interpretability

Access to Document

10.1109/COMPSAC57700.2023.00158

Cite this

Cuzzocrea, A., Alahy Ratul, Q. E., Belmerabet, I., & Serra, E. (2023). An AI Framework for Modelling and Evaluating Attribution Methods in Enhanced Machine Learning Interpretability. In H. Shahriar, Y. Teranishi, A. Cuzzocrea, M. Sharmin, D. Towey, AKM. J. A. Majumder, H. Kashiwazaki, J.-J. Yang, M. Takemoto, N. Sakib, R. Banno, & S. I. Ahamed (Eds.), Proceedings - 2023 IEEE 47th Annual Computers, Software, and Applications Conference, COMPSAC 2023 (pp. 1030-1036). (Proceedings - International Computer Software and Applications Conference; Vol. 2023-June). IEEE Computer Society. https://doi.org/10.1109/COMPSAC57700.2023.00158

Cuzzocrea, Alfredo ; Alahy Ratul, Qudrat E. ; Belmerabet, Islam et al. / An AI Framework for Modelling and Evaluating Attribution Methods in Enhanced Machine Learning Interpretability. Proceedings - 2023 IEEE 47th Annual Computers, Software, and Applications Conference, COMPSAC 2023. editor / Hossain Shahriar ; Yuuichi Teranishi ; Alfredo Cuzzocrea ; Moushumi Sharmin ; Dave Towey ; AKM Jahangir Alam Majumder ; Hiroki Kashiwazaki ; Ji-Jiang Yang ; Michiharu Takemoto ; Nazmus Sakib ; Ryohei Banno ; Sheikh Iqbal Ahamed. IEEE Computer Society, 2023. pp. 1030-1036 (Proceedings - International Computer Software and Applications Conference).

@inproceedings{a0279cd3a3bd46d5bcad9ca0a6c16220,

title = "An AI Framework for Modelling and Evaluating Attribution Methods in Enhanced Machine Learning Interpretability",

abstract = "In this paper, we propose a general methodology for estimating the degree of the attribution methods precision and generality in machine learning interpretability. Additionally, we propose a technique to measure the attribution consistency between two attribution methods. In our experiments, we focus on the two well-known model agnostic attribution methods, SHAP and LIME, then we evaluate them on two real applications in the attack detection field. Our proposed methodology highlights the fact that both LIME and SHAP are lacking precision, generality, and consistency. Therefore, more inspection is needed in the attribution research field.",

keywords = "Artificial Intelligence, Intelligent AI Tools, Machine Learning Interpretability",

author = "Alfredo Cuzzocrea and {Alahy Ratul}, {Qudrat E.} and Islam Belmerabet and Edoardo Serra",

note = "Publisher Copyright: {\textcopyright} 2023 IEEE.; 47th IEEE Annual Computers, Software, and Applications Conference, COMPSAC 2023 ; Conference date: 26-06-2023 Through 30-06-2023",

year = "2023",

doi = "10.1109/COMPSAC57700.2023.00158",

language = "English",

series = "Proceedings - International Computer Software and Applications Conference",

publisher = "IEEE Computer Society",

pages = "1030--1036",

editor = "Hossain Shahriar and Yuuichi Teranishi and Alfredo Cuzzocrea and Moushumi Sharmin and Dave Towey and Majumder, {AKM Jahangir Alam} and Hiroki Kashiwazaki and Ji-Jiang Yang and Michiharu Takemoto and Nazmus Sakib and Ryohei Banno and Ahamed, {Sheikh Iqbal}",

booktitle = "Proceedings - 2023 IEEE 47th Annual Computers, Software, and Applications Conference, COMPSAC 2023",

}

Cuzzocrea, A, Alahy Ratul, QE, Belmerabet, I & Serra, E 2023, An AI Framework for Modelling and Evaluating Attribution Methods in Enhanced Machine Learning Interpretability. in H Shahriar, Y Teranishi, A Cuzzocrea, M Sharmin, D Towey, AKMJA Majumder, H Kashiwazaki, J-J Yang, M Takemoto, N Sakib, R Banno & SI Ahamed (eds), Proceedings - 2023 IEEE 47th Annual Computers, Software, and Applications Conference, COMPSAC 2023. Proceedings - International Computer Software and Applications Conference, vol. 2023-June, IEEE Computer Society, pp. 1030-1036, 47th IEEE Annual Computers, Software, and Applications Conference, COMPSAC 2023, Hybrid, Torino, Italy, 26/06/23. https://doi.org/10.1109/COMPSAC57700.2023.00158

An AI Framework for Modelling and Evaluating Attribution Methods in Enhanced Machine Learning Interpretability. / Cuzzocrea, Alfredo; Alahy Ratul, Qudrat E.; Belmerabet, Islam et al.
Proceedings - 2023 IEEE 47th Annual Computers, Software, and Applications Conference, COMPSAC 2023. ed. / Hossain Shahriar; Yuuichi Teranishi; Alfredo Cuzzocrea; Moushumi Sharmin; Dave Towey; AKM Jahangir Alam Majumder; Hiroki Kashiwazaki; Ji-Jiang Yang; Michiharu Takemoto; Nazmus Sakib; Ryohei Banno; Sheikh Iqbal Ahamed. IEEE Computer Society, 2023. p. 1030-1036 (Proceedings - International Computer Software and Applications Conference; Vol. 2023-June).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - An AI Framework for Modelling and Evaluating Attribution Methods in Enhanced Machine Learning Interpretability

AU - Cuzzocrea, Alfredo

AU - Alahy Ratul, Qudrat E.

AU - Belmerabet, Islam

AU - Serra, Edoardo

PY - 2023

Y1 - 2023

N2 - In this paper, we propose a general methodology for estimating the degree of the attribution methods precision and generality in machine learning interpretability. Additionally, we propose a technique to measure the attribution consistency between two attribution methods. In our experiments, we focus on the two well-known model agnostic attribution methods, SHAP and LIME, then we evaluate them on two real applications in the attack detection field. Our proposed methodology highlights the fact that both LIME and SHAP are lacking precision, generality, and consistency. Therefore, more inspection is needed in the attribution research field.

AB - In this paper, we propose a general methodology for estimating the degree of the attribution methods precision and generality in machine learning interpretability. Additionally, we propose a technique to measure the attribution consistency between two attribution methods. In our experiments, we focus on the two well-known model agnostic attribution methods, SHAP and LIME, then we evaluate them on two real applications in the attack detection field. Our proposed methodology highlights the fact that both LIME and SHAP are lacking precision, generality, and consistency. Therefore, more inspection is needed in the attribution research field.

KW - Artificial Intelligence

KW - Intelligent AI Tools

KW - Machine Learning Interpretability

UR - http://www.scopus.com/inward/record.url?scp=85168879630&partnerID=8YFLogxK

U2 - 10.1109/COMPSAC57700.2023.00158

DO - 10.1109/COMPSAC57700.2023.00158

M3 - Conference contribution

AN - SCOPUS:85168879630

T3 - Proceedings - International Computer Software and Applications Conference

SP - 1030

EP - 1036

BT - Proceedings - 2023 IEEE 47th Annual Computers, Software, and Applications Conference, COMPSAC 2023

A2 - Shahriar, Hossain

A2 - Teranishi, Yuuichi

A2 - Cuzzocrea, Alfredo

A2 - Sharmin, Moushumi

A2 - Towey, Dave

A2 - Majumder, AKM Jahangir Alam

A2 - Kashiwazaki, Hiroki

A2 - Yang, Ji-Jiang

A2 - Takemoto, Michiharu

A2 - Sakib, Nazmus

A2 - Banno, Ryohei

A2 - Ahamed, Sheikh Iqbal

PB - IEEE Computer Society

T2 - 47th IEEE Annual Computers, Software, and Applications Conference, COMPSAC 2023

Y2 - 26 June 2023 through 30 June 2023

ER -

Cuzzocrea A, Alahy Ratul QE, Belmerabet I, Serra E. An AI Framework for Modelling and Evaluating Attribution Methods in Enhanced Machine Learning Interpretability. In Shahriar H, Teranishi Y, Cuzzocrea A, Sharmin M, Towey D, Majumder AKMJA, Kashiwazaki H, Yang JJ, Takemoto M, Sakib N, Banno R, Ahamed SI, editors, Proceedings - 2023 IEEE 47th Annual Computers, Software, and Applications Conference, COMPSAC 2023. IEEE Computer Society. 2023. p. 1030-1036. (Proceedings - International Computer Software and Applications Conference). doi: 10.1109/COMPSAC57700.2023.00158

An AI Framework for Modelling and Evaluating Attribution Methods in Enhanced Machine Learning Interpretability

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this