EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

4046. Enhancing Environmental Impact Assessment Procedures in Chile: Multi-Label Transformer Models for Predicting State Agencies Participation

Invited abstract in session WA-31: Analytics for Decision Making, stream Analytics.

Wednesday, 8:30-10:00
Room: 046 (building: 208)

Authors (first author is the speaker)

1. Alonso Leal
University of los Andes
2. Carla Vairetti
Universidad de los Andes
3. Sebastian Maldonado
Department of Management Control and Information Systems, University of Chile

Abstract

In Chile, the Environmental Impact Service (SEA) mandates assessments for industrial projects falling under specific categories, necessitating either an Environmental Impact Study (EIA) or an Environmental Impact Declaration (DIA). Central to the evaluation process is the invitation of the correct state administration agencies for participation. This investigation presents a novel approach leveraging pre-trained transformer models, tailored to predict the involvement of various state bodies in project assessments. Through the analysis of text within the project's primary form of declaration, we curated a dataset capturing participation patterns. Secondly, we fine-tuned the transformers to yield multi-label predictions, this was done by the modification of the final transformer layer, which facilitated the output of probabilistic estimates for each agency's participation. To address class imbalance in the dataset, we integrated class weights techniques into the loss function during model training. The experimental results showcase the effectiveness of this approach in accurately predicting multi-label participation, thereby enhancing the decision-making process surrounding environmental impact assessments. This research contributes to refining the SEAs procedures, ensuring comprehensive participation and evaluations of industrial projects by state agencies.

Keywords

Status: accepted

Back to the list of papers