AI time series forecasting

Chronos is a series of pre-trained time series prediction models based on language model architecture. The time series is converted into a sequence of tokens by scaling and quantization, and then a language model is trained using cross-entropy loss. After training, probabilistic predictions are obtained by sampling multiple future trajectories given historical context. Chronos models have been trained on a large amount of publicly available time series data and synthetic data generated using Gaussian processes.