Reports from University of Central Florida Highlight Recent Findings in Risk Management (Crash Data Augmentation Using Variational Autoencoder)

2021 MAR 29 (NewsRx) -- By a News Reporter-Staff News Editor at Insurance Daily News -- Researchers detail new data in Risk Management. According to news reporting from Orlando, Florida, by NewsRx journalists, research stated, “In this paper, we present a data augmentation technique to reproduce crash data. The dataset comprising crash and non-crash events are extremely imbalanced.”

The news correspondents obtained a quote from the research from the University of Central Florida, “For instance, the dataset used in this paper consists of only 625 crash events for over 6.5 million non-crash events. Thus, learning algorithms tend to perform poorly on these datasets. We have used variational autoencoder to encode all the events into a latent space. After training, the model could successfully separate crash and non-crash events. To generate data, we sampled from the latent space containing crash data. The generated data was compared with the real data from different statistical aspects. t-Test, Levene-test and Kolmogrove Smirnov test showed that the generated data was statistically similar to the real data. It was also compared to some of the minority oversampling techniques like SMOTE and ADASYN as well as the GAN framework for generating data. Crash prediction models based on Logistic Regression (LR), Support Vector Machine (SVM) and Artificial Neural Network (ANN) were used to compare the generated data from the different oversampling techniques. Overall, variational autoencoder (VAE) showed excellent results compared to the other data augmentation methods. Specificity is improved by 8% and 4% for VAE-LR and VAE-SVM respectively when compared to SMOTE while the sensitivity is improved by 6% and 5% when compared to ADASYN.”

According to the news reporters, the research concluded: “Moreover, VAE generated data also helps to overcome the overfitting problem in SMOTE and ADASYN since there is flexibility in choosing the decision boundary.”

This research has been peer-reviewed.

For more information on this research see: Crash Data Augmentation Using Variational Autoencoder. Accident Analysis & Prevention, 2021;151:105950. Accident Analysis & Prevention can be contacted at: Pergamon-elsevier Science Ltd, The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, England. (Elsevier - www.elsevier.com; Accident Analysis & Prevention - http://www.journals.elsevier.com/accident-analysis-and-prevention/)

Our news journalists report that additional information may be obtained by contacting Zubayer Islam, University of Central Florida, Dept. of Civil Environmental & Construction Engineering, Orlando, FL 32816, United States. Additional authors for this research include Mohamed Abdel-Aty, Qing Cai and Jinghui Yuan.

The direct object identifier (DOI) for that additional information is: https://doi.org/10.1016/j.aap.2020.105950. This DOI is a link to an online electronic document that is either free or for purchase, and can be your direct source for a journal article and its citation.

(Our reports deliver fact-based news of research and discoveries from around the world.)