The field of time series forecasting and analysis is rapidly evolving, with a focus on developing innovative methods to improve prediction accuracy and efficiency. Recent research has explored the integration of large language models (LLMs) with time series data, enabling the incorporation of contextual information and semantic knowledge into forecasting models. Another significant direction is the development of frameworks that combine multiple approaches, such as contrastive and generative methods, to leverage their complementary strengths. Additionally, there is a growing interest in applying time series analysis to various domains, including IoT, finance, and healthcare, where accurate forecasting and anomaly detection are crucial. Noteworthy papers in this area include QuiZSF, which proposes a lightweight and modular framework for zero-shot time series forecasting, and TALON, which enhances LLM-based forecasting by modeling temporal heterogeneity and enforcing semantic alignment. Furthermore, papers like TempOpt and AIS-LLM demonstrate the effectiveness of novel approaches in learning alarm relations and predicting maritime trajectories, respectively.
Advances in Time Series Forecasting and Analysis
Sources
Fractal Language Modelling by Universal Sequence Maps (USM)
QuiZSF: An efficient data-model interaction framework for zero-shot time-series forecasting
TLCCSP: A Scalable Framework for Enhancing Time Series Forecasting with Time-Lagged Cross-Correlations
Adapting LLMs to Time Series Forecasting via Temporal Heterogeneity Modeling and Semantic Alignment
Keyword-Centric Prompting for One-Shot Event Detection with Self-Generated Rationale Enhancements
Semantic-Enhanced Time-Series Forecasting via Large Language Models
AIS-LLM: A Unified Framework for Maritime Trajectory Prediction, Anomaly Detection, and Collision Risk Assessment with Explainable Forecasting
Hexagonal Picture Scanning Automata
TempOpt -- Unsupervised Alarm Relation Learning for Telecommunication Networks
TiMoE: Time-Aware Mixture of Language Experts
Fre-CW: Targeted Attack on Time Series Forecasting using Frequency Domain Loss
From Values to Tokens: An LLM-Driven Framework for Context-aware Time Series Forecasting via Symbolic Discretization
A Unified Contrastive-Generative Framework for Time Series Classification
DeepFeatIoT: Unifying Deep Learned, Randomized, and LLM Features for Enhanced IoT Time Series Sensor Data Classification in Smart Industries
TimeMKG: Knowledge-Infused Causal Reasoning for Multivariate Time Series Modeling
Anomaly Detection for IoT Global Connectivity
rETF-semiSL: Semi-Supervised Learning for Neural Collapse in Temporal Data
A Transformer-Based Approach for DDoS Attack Detection in IoT Networks