Decision Tree Extraction using Trained Neural Network

Nikola Vasilev, Zheni Mincheva, Ventsislav Nikolov

2020

Abstract

This paper introduces an algorithm for the extraction of rules from trained neural network. One of the main disadvantages of neural networks is their presumed complexity and people’s inability to fully comprehend their underlying logic. Their black box nature deems them useless in cases where the process of classification is important and must be presented in an observable and understandable way. The described algorithm extracts rules from a trained neural network and presents them in a form easily interpretable to humans. The paper demonstrates different approaches of rule extraction. Extracted rules explain and illustrate the network’s decision-making process. Rules can also be observed in the form of a tree. The presented algorithm generates rules by changing the input data and classifying them using the Reverse Engineering approach. After processing the data, the algorithm can use different approaches for creating the rules.

Paper Citation

in Harvard Style

Vasilev N., Mincheva Z. and Nikolov V. (2020). Decision Tree Extraction using Trained Neural Network.In Proceedings of the 9th International Conference on Smart Cities and Green ICT Systems - Volume 1: SMARTGREENS, ISBN 978-989-758-418-3, pages 194-200. DOI: 10.5220/0009351801940200

in Bibtex Style

@conference{smartgreens20,
author={Nikola Vasilev and Zheni Mincheva and Ventsislav Nikolov},
title={Decision Tree Extraction using Trained Neural Network},
booktitle={Proceedings of the 9th International Conference on Smart Cities and Green ICT Systems - Volume 1: SMARTGREENS,},
year={2020},
pages={194-200},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0009351801940200},
isbn={978-989-758-418-3},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 9th International Conference on Smart Cities and Green ICT Systems - Volume 1: SMARTGREENS,
TI - Decision Tree Extraction using Trained Neural Network
SN - 978-989-758-418-3
AU - Vasilev N.
AU - Mincheva Z.
AU - Nikolov V.
PY - 2020
SP - 194
EP - 200
DO - 10.5220/0009351801940200