PRIN 2022 Project
Artificial Intelligence-based Early Warning Tools
Research project funded under PRIN 2022
The project
Artificial Intelligence-based Early Warning Tools (AI-EWT)
The project critically and prospectively analyzed the early warning measures provided by the Italian Crisis and Insolvency Code and Directive (EU) 2019/1023, with specific attention to the needs and vulnerabilities of SMEs.
Research activities were carried out by two operational units: Unit 1 at the University of Campania “Luigi Vanvitelli” (Department of Law) and Unit 2 at the University of Naples Parthenope (Department of Business and Quantitative Studies).
Research project funded under PRIN 2022
Objectives
Research objectives
The project examined weaknesses and possible limitations in the application of the early warning measures established by the Italian Code of Business Crisis and Insolvency and Directive (EU) 2019/1023, with particular regard to SMEs.
The research combined legal analysis with economic-statistical methods to compare statutory indicators with quantitative and qualitative signs of business crisis and to formulate a regulatory proposal in the form of a “Model Law”.
Methodology
The use of artificial intelligence
Traditional machine-learning models are effective for credit scoring, but when used to predict insolvency, they tend to generate a high number of false positives. In other words, some companies are classified as being at risk of insolvency even though they do not subsequently fail.
Current models show significant margins of error in predicting insolvency.
Methodology
The Developed Predictive Model
The experimental workflow compares logistic regression, a feed-forward network without sequences and a multi-input LSTM network. The latter combines a three-year sequence, four static variables and regional information.
The target is named allert in the script; its construction is external to the published file. The page therefore describes the pipeline and temporal evaluations without assigning probabilities or horizons not defined in the code.
time_features_lstm_2020 <- layer_input(shape = c(3, 51), dtype = 'float32', name = "time_features_lstm_2020")
bilancio_lstm_2020 <- time_features_lstm_2020 %>%
layer_lstm(units = 64) %>%
layer_dropout(0.2)Project results
Regulatory proposal
As the final outcome of the research project, a comprehensive regulatory proposal in the form of a “Model Law” was formulated, aimed at introducing innovative criteria for early identification of crisis situations and more efficient and proportionate alert activation procedures, amending Leg. Decree 14/2019 with the introduction of Art. 5-ter and a quarterly automated predictive test.
In Article 3, paragraph 4 of Leg. Decree 14/2019, after letter d) the following is inserted: "e) the reporting of a crisis probability resulting from the predictive test..."
"An automated predictive test on the probability of a business crisis based on artificial intelligence tools is available on the national online platform referred to in Art. 13..."
Article 1 Amendments to Legislative Decree 14/2019
1. In Article 3(4) of Legislative Decree no. 14 of 12 January 2019, the following point is inserted after point (d): “(e) notification of a probability of crisis resulting from the predictive test on the probability of business crisis referred to in Article 5-ter”.
2. After Article 5-bis, the following is inserted: “Article 5-ter — Artificial intelligence test on business solvency. The national online platform referred to in Article 13 provides a predictive test on the probability of business crisis based on artificial intelligence tools. The test is run automatically every quarter...”
Article 2 — Implementing and financial provisions. The Ministry of Enterprises and Made in Italy shall issue a decree containing the provisions implementing the measures laid down in Article 1.
Publications
Project publications
Book and scientific contributions published within the AI-EWT project.
Among the publications: La gestione iniziale della crisi. L’intelligenza artificiale nella rilevazione tempestiva della crisi e gli strumenti preconcorsuali di risanamento (The initial management of the crisis. Artificial intelligence in early crisis detection and pre-insolvency restructuring tools), edited by Mario Campobasso and Maria Consiglia di Martino, Cacucci Editore.
Collective volume · 31 January 2025
Conferences
Conferences archive and initiatives
Events, workshops, publications and dissemination activities of the AI-EWT project.
Research impact
Impact of the Research
The introduction of more sophisticated predictive indicators can enable greater timeliness and accuracy in identifying difficulties, increase recovery chances, and foster the adoption of adequate organizational, administrative, and accounting structures, especially for SMEs.
Quantitative impact