Variational AutoEncoder for synthetic insurance data
This article explores the application of Variational AutoEncoders (VAEs) to insurance data. Previous research has demonstrated the successful implementation of generative models, especially VAEs, across various domains, such as image recognition, text classification, and recommender systems. However...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2024-12-01
|
| Series: | Intelligent Systems with Applications |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2667305324001297 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850109317588451328 |
|---|---|
| author | Charlotte Jamotton Donatien Hainaut |
| author_facet | Charlotte Jamotton Donatien Hainaut |
| author_sort | Charlotte Jamotton |
| collection | DOAJ |
| description | This article explores the application of Variational AutoEncoders (VAEs) to insurance data. Previous research has demonstrated the successful implementation of generative models, especially VAEs, across various domains, such as image recognition, text classification, and recommender systems. However, their application to insurance data, particularly to heterogeneous insurance portfolios with mixed continuous and discrete attributes, remains unexplored. This study introduces novel insights into utilising VAEs for unsupervised learning tasks in actuarial science, including dimension reduction and synthetic data generation. We propose a VAE model with a quantile transformation for continuous (latent) variables, a reconstruction loss that combines categorical cross-entropy and mean squared error, and a KL divergence-based regularisation term. Our VAE model’s architecture circumvents the need to pre-train and fine-tune a neural network to encode categorical variables into n-dimensional representative vectors within a continuous vector space of dimension Rn. We assess our VAE’s ability to reconstruct complex insurance data and generate synthetic insurance policies using a motor portfolio. Our experimental results and analysis highlight the potential of VAEs in addressing challenges related to data availability in the insurance industry. |
| format | Article |
| id | doaj-art-a7dc249cff694ebda650880954d12a41 |
| institution | OA Journals |
| issn | 2667-3053 |
| language | English |
| publishDate | 2024-12-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Intelligent Systems with Applications |
| spelling | doaj-art-a7dc249cff694ebda650880954d12a412025-08-20T02:38:06ZengElsevierIntelligent Systems with Applications2667-30532024-12-012420045510.1016/j.iswa.2024.200455Variational AutoEncoder for synthetic insurance dataCharlotte Jamotton0Donatien Hainaut1Corresponding author.; Institute of Statistics, Biostatistics and Actuarial Sciences (ISBA) of the Université catholique de Louvain (UCLouvain), Voie du Roman Pays 20/L1.04.01, Louvain-la-Neuve, 1348, BelgiumInstitute of Statistics, Biostatistics and Actuarial Sciences (ISBA) of the Université catholique de Louvain (UCLouvain), Voie du Roman Pays 20/L1.04.01, Louvain-la-Neuve, 1348, BelgiumThis article explores the application of Variational AutoEncoders (VAEs) to insurance data. Previous research has demonstrated the successful implementation of generative models, especially VAEs, across various domains, such as image recognition, text classification, and recommender systems. However, their application to insurance data, particularly to heterogeneous insurance portfolios with mixed continuous and discrete attributes, remains unexplored. This study introduces novel insights into utilising VAEs for unsupervised learning tasks in actuarial science, including dimension reduction and synthetic data generation. We propose a VAE model with a quantile transformation for continuous (latent) variables, a reconstruction loss that combines categorical cross-entropy and mean squared error, and a KL divergence-based regularisation term. Our VAE model’s architecture circumvents the need to pre-train and fine-tune a neural network to encode categorical variables into n-dimensional representative vectors within a continuous vector space of dimension Rn. We assess our VAE’s ability to reconstruct complex insurance data and generate synthetic insurance policies using a motor portfolio. Our experimental results and analysis highlight the potential of VAEs in addressing challenges related to data availability in the insurance industry.http://www.sciencedirect.com/science/article/pii/S2667305324001297AutoencoderVariational inferenceSynthetic data generationHeterogeneous insurance dataDimension reduction |
| spellingShingle | Charlotte Jamotton Donatien Hainaut Variational AutoEncoder for synthetic insurance data Intelligent Systems with Applications Autoencoder Variational inference Synthetic data generation Heterogeneous insurance data Dimension reduction |
| title | Variational AutoEncoder for synthetic insurance data |
| title_full | Variational AutoEncoder for synthetic insurance data |
| title_fullStr | Variational AutoEncoder for synthetic insurance data |
| title_full_unstemmed | Variational AutoEncoder for synthetic insurance data |
| title_short | Variational AutoEncoder for synthetic insurance data |
| title_sort | variational autoencoder for synthetic insurance data |
| topic | Autoencoder Variational inference Synthetic data generation Heterogeneous insurance data Dimension reduction |
| url | http://www.sciencedirect.com/science/article/pii/S2667305324001297 |
| work_keys_str_mv | AT charlottejamotton variationalautoencoderforsyntheticinsurancedata AT donatienhainaut variationalautoencoderforsyntheticinsurancedata |