Scholarlink

The synthetic generation of data using artificial intelligence has emerged as a strategic response to data scarcity, confidentiality, and imbalance. This study explores whether a Generative Adversarial Network (GAN) can generate synthetic daily sales data for Chilean SMEs that preserves statistical distributions and temporal patterns for sales forecasting. The methodology trains a tailored GAN on anonymized datasets and evaluates the generated data using distributional similarity, Dynamic Time Warping (DTW), and regression-based predictive performance. Results show that combining real and synthetic data improves forecasting accuracy while maintaining privacy. The study contributes to AI-driven data augmentation by adapting GANs to a local context with limited data availability, with a systematic evaluation framework, and by demonstrating the potential of synthetic datasets to support inclusive financing and digital transformation in underrepresented Latin American ecosystems.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Synthetic Sales Data Generation for Chilean SMEs Using GANs

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • David Ruete,
  • Marcia Durán-Riveros,
  • Sebastián Maldonado-Durán,
  • Alejandro Caroca,
  • Carla Taramasco,
  • Danilo Leal,
  • Nicolás Caselli,
  • Marcelo Reyes,
  • Omar Salinas,
  • Diego Mellado,
  • Jean Paul Maidana

摘要

The synthetic generation of data using artificial intelligence has emerged as a strategic response to data scarcity, confidentiality, and imbalance. This study explores whether a Generative Adversarial Network (GAN) can generate synthetic daily sales data for Chilean SMEs that preserves statistical distributions and temporal patterns for sales forecasting. The methodology trains a tailored GAN on anonymized datasets and evaluates the generated data using distributional similarity, Dynamic Time Warping (DTW), and regression-based predictive performance. Results show that combining real and synthetic data improves forecasting accuracy while maintaining privacy. The study contributes to AI-driven data augmentation by adapting GANs to a local context with limited data availability, with a systematic evaluation framework, and by demonstrating the potential of synthetic datasets to support inclusive financing and digital transformation in underrepresented Latin American ecosystems.