Abstract:Time-series forecasting is an important problem across a wide range of domains. Designing accurate and prompt forecasting algorithms is a non-trivial task, as temporal data that arise in real applications often involve both non-linear dynamics and linear dependencies, and always have some mixtures of sequential and periodic patterns, such as daily, weekly repetitions, and so on. At this point, however, most recent deep models often use Recurrent Neural Networks (RNNs) to capture these temporal patterns, which is hard to parallelize and not fast enough for real-world applications especially when a huge amount of user requests are coming. Recently, CNNs have demonstrated significant advantages for sequence modeling tasks over the de-facto RNNs, while providing high computational efficiency due to the inherent parallelism. In this work, we propose HyDCNN, a novel hybrid framework based on fully Dilated CNN for time-series forecasting tasks. The core component in HyDCNN is a proposed hybrid module, in which our proposed position-aware dilated CNNs are utilized to capture the sequential non-linear dynamics and an autoregressive model is leveraged to capture the sequential linear dependencies. To further capture the periodic temporal patterns, a novel hop scheme is introduced in the hybrid module. HyDCNN is then composed of multiple hybrid modules to capture the sequential and periodic patterns. Each of these hybrid modules targets on either the sequential pattern or one kind of periodic patterns. Extensive experiments on five real-world datasets have shown that the proposed HyDCNN is better compared with state-of-the-art baselines and is at least 200% better than RNN baselines. The datasets and source code will be published in Github to facilitate more future work.

Hierarchical Time Series Forecasting Based on Temporal Convolution and Attention Mechanism

Hidformer: Hierarchical Dual-Tower Transformer Using Multi-Scale Mergence for Long-Term Time Series Forecasting

The Temperature Prediction of Hydro-generating Units based on Temporal Convolutional Network and Recurrent Neural Network

TATCN: Time Series Prediction Model Based on Time Attention Mechanism and TCN

Enhancing Time Series Forecasting: A Hierarchical Transformer with Probabilistic Decomposition Representation

Hierarchical Classification Auxiliary Network for Time Series Forecasting

Forecasting Hierarchical Time Series

Multi-task Learning Method for Hierarchical Time Series Forecasting

A Multi-Hierarchical Attention-Based Prediction Method on Time Series with Spatio-Temporal Context among Variables

Efficient Forecasting of Large Scale Hierarchical Time Series via Multilevel Clustering

Parallel spatio-temporal attention-based TCN for multivariate time series prediction

Temporal Chain Network with Intuitive Attention Mechanism for Long-Term Series Forecasting.

Multi-Horizon Time Series Forecasting with Temporal Attention Learning

Flow-Based End-to-End Model for Hierarchical Time Series Forecasting via Trainable Attentive-Reconciliation

Multivariate Temporal Convolutional Network: A Deep Neural Networks Approach for Multivariate Time Series Forecasting

End-to-End Modeling of Hierarchical Time Series Using Autoregressive Transformer and Conditional Normalizing Flow-based Reconciliation

High-efficiency Chaotic Time Series Prediction Based on Time Convolution Neural Network

DDTCN: Decomposed dimension time-domain convolutional neural network along spatial dimensions for multiple long-term series forecasting

Probabilistic Forecasting with Temporal Convolutional Neural Network

Modeling Temporal Patterns with Dilated Convolutions for Time-Series Forecasting.

End-to-End Modeling Hierarchical Time Series Using Autoregressive Transformer and Conditional Normalizing Flow based Reconciliation