A Novel Tensor Causal Convolution Network Model for Highly-Accurate Representation to Spatio-Temporal Data

Spatio-temporal data like a Dynamically Weighted Directed Network (DWDN) are ubiquitous in bigdata applications like an intelligent recommender system. They commonly illustrate the complex yet dynamic interactions between tremendous nodes, as well as contain rich knowledge regarding the involved nodes’ behavioral patterns with time dynamics. On the other hand, since the nodes constantly increase as the time accumulates, a DWDN becomes High-Dimensional and Incomplete (HDI) due to the limited interactions, widely-spread time slots and huge node count. Moreover, the inner temporal-spatio patterns exhibit strong nonlinearity, making it very difficult to grasp them from HDI data. To address this critical issue, this paper proposes a novel Tensor Causal Convolution Network (TCCN) model with three-fold ideas: a) innovatively building a feedforward tensor neural network with the incorporation of the Latent Factorization of Tensors (LFT) principle to model complex nonlinear interactions in an HDI DWDN efficiently; b) establishing a Tensor Causal Convolution (TCC) structure, which is able to accurately fuse the time-varying information from node interactions with high scalability; and c) developing a neighborhood regularization scheme to boost the local structural representation, thus capturing spatial dependencies among tremendous nodes. Extensively experimental results on ten real DWDNs from real bigdata applications evidently demonstrate that our proposed TCCN model outperforms several state-of-the-art models in both representation learning accuracy and convergence ability. It provides a highly-efficient representation learning approach for diverse bigdata applications. Note to Practitioners—In modern industrial systems such as recommendation platforms, IoT device networks, and financial transaction monitoring systems, tremendous and temporal interactions emerge among numerous nodes over time. This leads to large-scale dynamic networks that are high-dimensional and incomplete (HDI), making it difficult for traditional methods to extract meaningful behavioral patterns. This paper introduces an efficient method for modeling and representing node relationships and temporal patterns in such a network. The proposed approach, namely the TCCN model, is able to automatically fuse the time-varying features to preserve local structural dependencies, hence significantly improving the interpretation of user behavior, device states, and transaction flows. Experimental results across several real-world industrial scenarios demonstrate TCCN’s superior performance in both representation accuracy and model convergence. In future research, we plan to explore the generative modeling frameworks to improve TCCN’s generalization and adaptability to complex temporal dynamics. In addition, TCNN also holds promise for broader applications such as intelligent transportation systems, industrial process monitoring, and energy management in data-sparse automation contexts.