Petrônio Cândido Lima Silva
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Petrônio Cândido Lima Silva has a total of 82 co-authors. Here are the top co-authors by number of shared publications:
- Frederico Gadelha Guimarães
Universidade Federal de Minas Gerais
31 shared publications - Omid Orang
Federal University of Minas Gerais
16 shared publications - Marcos Antonio Alves
FITec Technological Innovations, Belo Horizonte 30140-150, Brazil
6 shared publications - Patrícia de Oliveira Lucas
Federal Institute of Northern Minas Gerais Salinas MG Brazil
6 shared publications - Fabricio Javier Erazo-Costa
Graduate Program in Electrical Engineering, Universidade Federal de Minas Gerais, Belo Horizonte, Brazil
4 shared publications - Felipe Augusto Rocha da Silva
Graduate Program in Electrical Engineering Universidade Federal de Minas Gerais Av. Antônio Carlos 6627 31270-901 Belo Horizonte MG Brazil
4 shared publications
Latest Publications
- 1 October 2026
- Published by Springer Nature
- p. 263-277
Open-Set Zero-Day Release Authorship Attribution of Code for Digital Forensics
- 1 October 2026
- Published by Springer Nature
- p. 263-277
Open-Set Zero-Day Release Authorship Attribution of Code for Digital Forensics
- 4 June 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 331
Spatiotemporal modeling with graph neural cellular automata for modular traffic forecasting
- 4 June 2026
- Published by Elsevier in Expert Systems with Applications
- Vol. 331
Spatiotemporal modeling with graph neural cellular automata for modular traffic forecasting
- Open Access
- 13 March 2026
- Published by Elsevier in SSRN Electronic Journal
Spatiotemporal modeling with Graph Neural Cellular Automata for Modular Traffic Forecasting
Spatiotemporal forecasting (STF) plays a critical role in modeling phenomena characterized by intrinsic temporal and spatial dependencies, with urban traffic flow forecasting representing one of its most prominent applications. Despite advances in data availability from smart city infrastructures, existing deep learning approaches, particularly those based on graph neural networks (GNNs), often rely on monolithic architectures that lack flexibility in handling heterogeneous or evolving graph structures. Moreover, training such models remains computationally demanding, and effective transfer learning across diverse graph configurations remains an open research challenge. Thus, this work introduces Graph Neural Cellular Automata (GNCA), a modular and scalable framework that integrates graph-based representations with local interaction principles inspired by cellular automata. The proposed model employs a novel multi-embedding strategy that combines temporal, spatial, and value embeddings to capture complex spatiotemporal dependencies, alongside a flexible neighborhood tokenizer capable of adapting to arbitrary graph topologies. Evaluated on established traffic forecasting benchmarks, GNCA demonstrates competitive performance against state-of-the-art methods, offering a versatile and efficient paradigm for spatiotemporal time-series forecasting.
- 13 March 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessSpatiotemporal modeling with Graph Neural Cellular Automata for Modular Traffic Forecasting
Spatiotemporal forecasting (STF) plays a critical role in modeling phenomena characterized by intrinsic temporal and spatial dependencies, with urban traffic flow forecasting representing one of its most prominent applications. Despite advances in data availability from smart city infrastructures, existing deep learning approaches, particularly those based on graph neural networks (GNNs), often rely on monolithic architectures that lack flexibility in handling heterogeneous or evolving graph structures. Moreover, training such models remains computationally demanding, and effective transfer learning across diverse graph configurations remains an open research challenge. Thus, this work introduces Graph Neural Cellular Automata (GNCA), a modular and scalable framework that integrates graph-based representations with local interaction principles inspired by cellular automata. The proposed model employs a novel multi-embedding strategy that combines temporal, spatial, and value embeddings to capture complex spatiotemporal dependencies, alongside a flexible neighborhood tokenizer capable of adapting to arbitrary graph topologies. Evaluated on established traffic forecasting benchmarks, GNCA demonstrates competitive performance against state-of-the-art methods, offering a versatile and efficient paradigm for spatiotemporal time-series forecasting.
- 30 January 2026
- Published by Springer Nature
- p. 424-438
Enhanced Forecasting Model Using Transformers, Extended Long-Short-Term Memory, and Randomized Fuzzy Cognitive Maps
- 30 January 2026
- Published by Springer Nature
- p. 424-438
Enhanced Forecasting Model Using Transformers, Extended Long-Short-Term Memory, and Randomized Fuzzy Cognitive Maps
- 11 November 2025
- Published by Springer Nature in Evolving Systems
- Vol. 16 (4) , 1-10
Automated forecasting model leveraging large reservoirs of randomized fuzzy cognitive maps
- 11 November 2025
- Published by Springer Nature in Evolving Systems
- Vol. 16 (4) , 1-10
Automated forecasting model leveraging large reservoirs of randomized fuzzy cognitive maps
- Open Access
Causal Graph Fuzzy LLMs: A First Introduction and Applications in Time Series Forecasting
In recent years, the application of Large Language Models (LLMs) to time series forecasting (TSF) has garnered significant attention among researchers. This study presents a new frame of LLMs named CGF-LLM using GPT-2 combined with fuzzy time series (FTS) and causal graph to predict multivariate time series, marking the first such architecture in the literature. The key objective is to convert numerical time series into interpretable forms through the parallel application of fuzzification and causal analysis, enabling both semantic understanding and structural insight as input for the pretrained GPT-2 model. The resulting textual representation offers a more interpretable view of the complex dynamics underlying the original time series. The reported results confirm the effectiveness of our proposed LLM-based time series forecasting model, as demonstrated across four different multivariate time series datasets. This initiative paves promising future directions in the domain of TSF using LLMs based on FTS.
Open AccessCausal Graph Fuzzy LLMs: A First Introduction and Applications in Time Series Forecasting
In recent years, the application of Large Language Models (LLMs) to time series forecasting (TSF) has garnered significant attention among researchers. This study presents a new frame of LLMs named CGF-LLM using GPT-2 combined with fuzzy time series (FTS) and causal graph to predict multivariate time series, marking the first such architecture in the literature. The key objective is to convert numerical time series into interpretable forms through the parallel application of fuzzification and causal analysis, enabling both semantic understanding and structural insight as input for the pretrained GPT-2 model. The resulting textual representation offers a more interpretable view of the complex dynamics underlying the original time series. The reported results confirm the effectiveness of our proposed LLM-based time series forecasting model, as demonstrated across four different multivariate time series datasets. This initiative paves promising future directions in the domain of TSF using LLMs based on FTS.
- Open Access
- 24 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 56180-56194
Time Series Classification Using Federated Convolutional Neural Networks and Image-Based Representations
Time series classification (TSC) has numerous applications across various domains. This research introduces a federated hybrid TSC method that combines image-based time series representation techniques with Convolutional Neural Networks (CNNs) in a decentralized framework. Specifically, three image representation techniques—Motif representation of numerical time series (NMotif), Motif representation of fuzzy time series (FTSMotif), and Polar—are utilized to convert raw data into image representations, which are then used as inputs for CNNs. Federated Learning (FL) enables collaborative training across distributed clients while ensuring data privacy. The method’s effectiveness is evaluated using 40 datasets from the UCR Archive. The results show that the FL model surpasses traditional baselines by up to 1.9% while enabling collaborative learning and preserving privacy. Although centralized state-of-the-art (SOTA) models such as Rocket and ResNet achieve higher overall accuracy, the proposed approach offers a competitive and practical alternative for real-world scenarios where centralized training is infeasible due to privacy constraints. These findings highlight the potential of our proposed federated image-based TSC as a scalable and privacy-aware solution for time series analysis.
- 24 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 56180-56194
Open AccessTime Series Classification Using Federated Convolutional Neural Networks and Image-Based Representations
Time series classification (TSC) has numerous applications across various domains. This research introduces a federated hybrid TSC method that combines image-based time series representation techniques with Convolutional Neural Networks (CNNs) in a decentralized framework. Specifically, three image representation techniques—Motif representation of numerical time series (NMotif), Motif representation of fuzzy time series (FTSMotif), and Polar—are utilized to convert raw data into image representations, which are then used as inputs for CNNs. Federated Learning (FL) enables collaborative training across distributed clients while ensuring data privacy. The method’s effectiveness is evaluated using 40 datasets from the UCR Archive. The results show that the FL model surpasses traditional baselines by up to 1.9% while enabling collaborative learning and preserving privacy. Although centralized state-of-the-art (SOTA) models such as Rocket and ResNet achieve higher overall accuracy, the proposed approach offers a competitive and practical alternative for real-world scenarios where centralized training is infeasible due to privacy constraints. These findings highlight the potential of our proposed federated image-based TSC as a scalable and privacy-aware solution for time series analysis.
- 10 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12 (12) , 21679-21690
A Multistep Multivariate Fuzzy-Based Time-Series Forecasting on Internet of Things Data
Multi-step ahead time series forecasting is essential in Internet of Things (IoT) applications in smart cities and smart homes to make accurate future predictions and precise decision-making. Thus, this study introduces a novel multiple-input single-output (MISO) forecasting method called Multi-Step Embedding-based Fuzzy Time Series (MS-EFTS), designed to predict high-dimensional non-stationary time series data. As a first-order approach, it employs a direct strategy that integrates an embedding transformation with a weighted multivariate FTS (WMVFTS) model. This combination allows for effective predictions over long-term horizons within low-dimensional, learned continuous representations. The effectiveness of the proposed MS-EFTS is assessed using three high-dimensional IoT time series in this investigation. The obtained results showcase the superior performance of the proposed method compared to some deep learning forecasting methods, including LSTM, BiLSTM, TCN, and CNN-LSTM, in terms of accuracy, parsimony, and efficiency.
- 10 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 12 (12) , 21679-21690
A Multistep Multivariate Fuzzy-Based Time-Series Forecasting on Internet of Things Data
Multi-step ahead time series forecasting is essential in Internet of Things (IoT) applications in smart cities and smart homes to make accurate future predictions and precise decision-making. Thus, this study introduces a novel multiple-input single-output (MISO) forecasting method called Multi-Step Embedding-based Fuzzy Time Series (MS-EFTS), designed to predict high-dimensional non-stationary time series data. As a first-order approach, it employs a direct strategy that integrates an embedding transformation with a weighted multivariate FTS (WMVFTS) model. This combination allows for effective predictions over long-term horizons within low-dimensional, learned continuous representations. The effectiveness of the proposed MS-EFTS is assessed using three high-dimensional IoT time series in this investigation. The obtained results showcase the superior performance of the proposed method compared to some deep learning forecasting methods, including LSTM, BiLSTM, TCN, and CNN-LSTM, in terms of accuracy, parsimony, and efficiency.
- 30 January 2025
- Published by Springer Nature
- p. 445-459
Traffic Forecasting Using Federated Randomized High-Order Fuzzy Cognitive Maps
- 30 January 2025
- Published by Springer Nature
- p. 445-459
Traffic Forecasting Using Federated Randomized High-Order Fuzzy Cognitive Maps