Granger-TSllm: Granger causality enhanced LLMs with residual-quantized tokenizer for multivariate time series forecasting.

Correctformer: A transformer architecture for correcting periodic drift in time-series forecasting.

PMTE-LLM:An LLM-based time series forecasting method using professional mechanism and training experience.

Hybrid-sequence self-learning model: Unsupervised anomaly detection and localization in multivariate time series

Iterative feedback-based time-series anomaly detection with adaptive diffusion models.

TimeCNN: Refining inscross-variable interaction on time point for time series forecasting.

Repetitive contrastive learning enhances Mamba's selectivity in time series prediction.

This data descriptor describes three datasets with hourly residential energy use data and building information data from apartments, cabins and single-family houses. The first dataset contains 29 files with energy use data from 29 apartment block condominiums, with hourly time series of both electricity and delivered heat. The second dataset includes 20 files from Risvollan borettslag in Trondheim, covering heating and electricity use at the level of entire apartment blocks and heating centrals, as well as sub-metered electricity data from 1058 individual apartments. The third dataset consists of 194 files with electricity smart meter data from single-family houses, cabins, and apartments equipped with Sikom home energy management systems. The energy use data have been collected in co-operation with distribution system operators and utility providers, directly from building owners as well as from Sikom with permission from their users. Information about the Sikom users have been collected through a user survey. Each dataset is organized into human readable comma-separated txt files with a common structure that includes building information data, hourly records of electricity and heating energy use, and weather variables, with time series durations from several months up to multiple years per building. The data are suitable for tasks such as residential building energy analysis, load profile generation, energy disaggregation, classification tasks, forecasting of energy use, demand flexibility analysis, energy system analysis and other modelling tasks.

Dual decomposition-enhanced integrated deep networks with bidirectional CNN and semi-supervised GRU for multivariate nonlinear time series forecasting

Magnetic resonance imaging (MRI) has limited capacity to visualise cortical bone due to its low proton density and short decay time. Recently developed ultrashort (UTE) and zero-echo time (ZTE) sequences enable bone imaging without ionising radiation. This study aims to identify the key technical parameters, advantages, and limitations of UTE and ZTE for musculoskeletal (MSK) imaging. JBI methodology was applied, and three databases (MEDLINE, EMBASE and CINAHL) were selected to identify articles published after 2005 (French-English). Keywordsand MeSH terms related to UTE, ZTE, MRI and MSK were used. Two independent reviewers screened titles, abstracts, and full texts. Disagreements were solved through consensus. From 671 articles, 18 met all criteria. UTE and ZTE were applied for spine (6/18), lower limb (4/18), head and neck (4/18), general bone (3/18) and shoulder (1/18) investigations. Eight articles suggest a very short repetition time (TR) (0.425-8 ms), three longer TR (100-1075 ms), six did not mention TR. All articles mentioned very short echo time (TE) (0.00-0.34 ms). Acquisition time ranged from 3 to 12 min. UTE and ZTE are based on radial acquisition, allowing to acquire the cortical bone signal and increasing fracture contrast, comparable to CT-scan. Acquisition time was the main disadvantage. UTE and ZTE sequences are promising for the MSK applications, offering good contrast for cortical bone evaluation. Further studies are necessary to assess the possibilities of AI tools as approaches to improve image quality and reduce acquisition time. UTE and ZTE sequences can be added to MSK MRI exams to improve fracture and cortical bone evaluation, allowing radiation-free imaging. ZTE's lower acoustic noise benefits anxious, paediatric, or dementia patients' MRI experience.

Hybrid multivariate time series prediction system based on PD-XGBoost-KRVFL

Spatial information bottleneck graph structure learning based multivariate time series prediction for industrial processes

KiST: Kernel improved spectral theory model for multivariate time series forecasting

Hierarchical prediction of irregular multivariate time series from a multi-granularity perspective

PrivTSAD-FedWGAN: A novel federated learning and WGAN framework for privacy-preserving multivariate time series anomaly detection

DDformer: Transformer with dynamic variable fusion and dynamic difference attention for multivariate time series long-term forecasting

AsynFormer: Transformer capturing asynchronous cross-variate dependencies for efficient multivariate time series forecasting

SFAFormer: Sampling Frequency-Aware Transformer Specialized for Unsupervised Anomaly Detection in Irregular Multivariate Time Series

A novel dynamic graph attention aggregation network for multivariate time series classification

HASPFormer: Advancing multivariate time-series forecasting with self-attention and stochastic pooling