Scholarlink

<p>Research in social community mining from social media faces challenges due to the constantly evolving nature of online networks. Deep learning methods often struggle with noisy data caused by sudden changes and fail to capture essential temporal information for community evolution. Additionally, data security concerns are frequently overlooked. This paper presents a novel probabilistic generative adversarial network (GAN) model designed for dynamic social community detection, addressing challenges posed by the evolving nature of networks and noisy data. Our approach integrates a hybrid probabilistic convolutional neural network (PCNN) and gated recurrent unit (GRU) architecture, capturing temporal community dynamics while preserving data confidentiality through deep generative masking techniques. The model generates synthetic time series data that effectively mimics real data while safeguarding sensitive information and improving forecasting accuracy. Experimental results demonstrate the model’s superior performance in terms of RMSE, MAPE, CE, CRPS*, and ARR, outperforming traditional deep learning models and providing valuable insights for decision-making in spatiotemporal community detection.</p>

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

Generating prediction clustering graph network analysis data from masked time series with GANs

  • 期刊论文

  • 发表日期：

    2025年2月28日

查看全文
  • Radhia Toujani,
  • Fatma Hendaoui,
  • Ahlem Ferchichi,
  • Mejda Chihaoui,
  • Aya Ferchichi,
  • Wided Bakari

摘要

Research in social community mining from social media faces challenges due to the constantly evolving nature of online networks. Deep learning methods often struggle with noisy data caused by sudden changes and fail to capture essential temporal information for community evolution. Additionally, data security concerns are frequently overlooked. This paper presents a novel probabilistic generative adversarial network (GAN) model designed for dynamic social community detection, addressing challenges posed by the evolving nature of networks and noisy data. Our approach integrates a hybrid probabilistic convolutional neural network (PCNN) and gated recurrent unit (GRU) architecture, capturing temporal community dynamics while preserving data confidentiality through deep generative masking techniques. The model generates synthetic time series data that effectively mimics real data while safeguarding sensitive information and improving forecasting accuracy. Experimental results demonstrate the model’s superior performance in terms of RMSE, MAPE, CE, CRPS*, and ARR, outperforming traditional deep learning models and providing valuable insights for decision-making in spatiotemporal community detection.