The automotive recall data search and its analysis applying machine learning

Maione, Bruno Fernandes; Kaminski, Paulo Carlos; Baraldi, Emilio Carlos

doi:10.1590/0103-6513.20220117

Bruno Fernandes Maione ^* * bruno.maione@usp.br

Universidade de São Paulo, São Paulo, SP, Brasil

https://orcid.org/0000-0002-8071-8228

Paulo Carlos Kaminski

Universidade de São Paulo, São Paulo, SP, Brasil

https://orcid.org/0000-0001-6917-7208

Emilio Carlos Baraldi

Universidade de São Paulo, São Paulo, SP, Brasil

https://orcid.org/0000-0001-7788-6442

* bruno.maione@usp.br

Thumbnail

Figure 1
Number of licensed new motor vehicles compared to number of motor vehicles involved in recall for the years 2011 to 2019. Source: Maione et al. (2021b)Maione, B. F., Kaminski, P. C., & Baraldi, E. C. (2021b). The different legislation of automotive recall and their implications for society. SAE International, 36, 1-12. http://dx.doi.org/10.4271/2020-36-0084.
http://dx.doi.org/10.4271/2020-36-0084... .

Thumbnail

Figure 2
Flowchart of machine learning techniques. Source: Adapted from Rafique & Velasco (2018)Rafique, D., & Velasco, L. (2018). Machine learning for network automation: overview, architecture, and applications [Invited Tutorial]. Journal of Optical Communications and Networking, 10(10), D126-D143. http://dx.doi.org/10.1364/JOCN.10.00D126.
http://dx.doi.org/10.1364/JOCN.10.00D126... and Géron (2019)Géron, A. (2019). Hands-on Machine Learning with Scikit-Learn, Keras & Tensor Flow: Concepts, tools, and techniques to build intelligent systems (2nd ed.). Canada: O’Reilly Media..

Thumbnail

Figure 3
Clustered data of Brazilian Recalls.

Thumbnail

Figure 4
Clustered data of USA Recalls.

Thumbnail

Figure 5
Data distribution of the elements used to build the model of classification.

Thumbnail

Figure 6
Data distribution of the elements used to build the model of classification.

Thumbnail

Figure 7
Regression results for both analyses with Brazilian data.

Thumbnail

Figure 8
Regression result for the first analysis with European Union data.

Thumbnail

Figure 9
Regression results for both analyses with USA data.

Thumbnail

Figure 10
Regression results for the three regions.

Thumbnail

Figure 3
Clustered data of Brazilian Recalls.

Thumbnail

Figure 4
Clustered data of USA Recalls.

Thumbnail

Figure 5
Data distribution of the elements used to build the model of classification.

Thumbnail

Figure 6
Data distribution of the elements used to build the model of classification.

Thumbnail

Figure 7
Regression results for both analyses with Brazilian data.

Thumbnail

Figure 8
Regression result for the first analysis with European Union data.

Thumbnail

Figure 9
Regression results for both analyses with USA data.

Thumbnail

Figure 10
Regression results for the three regions.

Thumbnail

Table 1
Comparison between the three regions.

Thumbnail

Table 2
Comparison between the three databases.

Thumbnail

Table 3
Approximate accuracy of the KNN Algorithm for each location.

Thumbnail

Table 4
Coefficients and parameters of the regression with Brazilian data.

Thumbnail

Table 5
Coefficients and parameters of the regression with European Union data.

Thumbnail

Table 6
Coefficients and parameters of the regression with USA data.

Thumbnail

Table 7
Coefficients and parameters of the three regions.

(1)

(2)

(3)

(4)

Region	Brazil	European Union	United States of America
Legislation	Constituição Federal Brasileira (Brazilian Federal Constitution)	European Parliament and the Council of the European Union	Federal Regulations of the United States of America
Law	Artigo 5º, inciso XXXII, da Constituição¸ Federal Brasileira (Article 5, Item XXXII, of the Brazilian Federal Constitution)	Directive 2001/95/EC	National Traffic and Motor Vehicle Safety Act of 1966 (15 U.S.C 1381)
Government agency	Ministério da Justiça (Justice Ministry)	European Commission	National Highway Traffic Safety Administration (NHTSA)
Other government agencies	Secretaria Nacional do Consumidor (National Consumer Secretariat)	National authorities of the members of the European Commission	United States Environmental Protection Agency (EPA)
Other government agencies	Grupo de Estudos Permanentes de Acidentes de Consumo (Permanent Study Group on Consumer Accidents)		Office of Vehicle Safety Research and the Office of Behavioral Safety Research
Access to recall information	Ministério da Justiça (Justice Ministry) – (Secretaria Nacional do Consumidor, 2020bSecretaria Nacional do Consumidor - SENACON. (2020b). Alertas de Recall. Retrieved in 2020, November 27, from http://portal.mj.gov.br/recall/ http://portal.mj.gov.br/recall/... )	Rapid Alert System for dangerous non-food products (RAPEX) - (European Commission, 2020European Commission (2020). Safety Gate: Rapid Alert System for dangerous non-food products - Search alerts. Retrieved in 2020, August 22, from https://ec.europa.eu/consumers/consumerssafety/safetyproducts/rapex/alerts/?event=main.searchlng=ensearchResults https://ec.europa.eu/consumers/consumers... )	U.S. Government recalls - (Recalls, 2018Recalls. (2018). Your online resource for recalls. Retrieved in 2018, July. 10, from www.recalls.gov www.recalls.gov... )

+	Brazil	European Union	United States of America
Recall Title	Present	Present	Present
Identification	Present	Present	Present
Risk Description	Present	Present	Present
Type of Product	Present	Present	Present
Manufacturer	Present	Present	Present
Country of Origin	Present	Present	Absent
Country of Export	Present	Present	Absent
Model	Present	Present	Present
Year of Manufacturer	Present	Present	Present
Number of Affected	Present	Absent	Present
Link to source	Present	Present	Absent
Date of Recall	Present	Present	Present
Type of Alert	Absent	Present	Present
Origin of Alert	Not Valid	Present	Not Valid
Measures taken	Absent	Present	Present
Affected Component	Present	Present	Present

KNN Algorithm	Accuracy
Brazil	70%
European Union (EU)	82%
United States of America (USA)	-

Regression - Brazil	Number of recalls	Number of affected vehicles
Regression Coefficient (β)	13.15	2.89E+5
Mean Squared Error (MSE)	988	5.76E+12
Coefficient of Determination (R²)	0.59	0.11

Regression - EU	Number of recalls	Number of affected vehicles*
Regression Coefficient (β)	42.11	1.48E+06
Mean Squared Error (MSE)	2831	5.43E+13
Coefficient of Determination (R²)	0.84	0.07

Regression - USA	Number of recalls	Number of affected vehicles
Regression Coefficient (β)	42.79	4.30E+06
Mean Squared Error (MSE)	5448	3.96E+14
Coefficient of Determination (R²)	0.73	0.28

Regression	Brazil		European Union		United States of America
Regression	Number of recalls	Number of affected vehicles	Number of recalls	Number of affected vehicles*	Number of recalls	Number of affected vehicles
Regression Coefficient (β)	13.15	2.89E+5	42.11	1.48E+06	42.79	4.30E+06
Mean Squared Error (MSE)	988	5.76E+12	2831	5.43E+13	5448	3.96E+14
Coefficient of Determination (R²)	0.59	0.11	0.84	0.07	0.73	0.28

[1] * bruno.maione@usp.br

Brasil

Brasil

The automotive recall data search and its analysis applying machine learning

Abstract

Paper aims

Originality

Research method

Main findings

Implications for theory and practice