81 repos across 4 sub-areas
Libraries, models, and frameworks for forecasting time series data using machine learning and deep learning approaches. The cluster spans foundational forecasting methods, neural network architectures for temporal prediction, and practical implementations across Python and R. Key repos include Lag-Llama and the Moirai family (foundational time-series forecasting models), alongside supporting tools for data science workflows in this domain.
Time Series Forecasting Models
32 repos
Libraries and models for predicting future values in sequential time-based data, with emphasis on deep learning approaches like transformers. The cluster centers on specialized forecasting architectures (particularly the Moirai and Granite TimeSeries models) alongside supporting tools for time series analysis and transformation. Repositories here span implementations, variants, and applications of modern neural forecasting methods.
Cluster 638380
25 repos
Time Series Forecasting & Foundation Models
21 repos
Machine learning approaches to time series prediction and forecasting, centered on large foundation models adapted for temporal data. The cluster includes the Chronos family of models—pretrained transformers fine-tuned for zero-shot and few-shot forecasting across diverse datasets—alongside supporting libraries and datasets for evaluating and implementing time series forecasting systems. This area bridges general-purpose deep learning with domain-specific temporal reasoning.
Time Series Forecasting & Deep Learning
3 repos
Python-based libraries and applications for time series analysis, forecasting, and prediction using machine learning and deep learning techniques. The cluster includes general-purpose forecasting frameworks like DARTS alongside specialized implementations for domains like glucose level prediction, as well as datasets and preprocessing pipelines for time series research. Practitioners here would find tools for building neural network models on temporal data, working with multivariate sequences, and evaluating forecast accuracy across different problem domains.