Cisco Time Series Model Technical Report

TL;DR

The Cisco Time Series Model is a novel foundation model for univariate forecasting that leverages multiresolution input to improve long-term prediction accuracy, trained on extensive data, and evaluated on multiple datasets.

Contribution

Introduces a multiresolution decoder-only time series model that enhances long context forecasting without sacrificing general performance.

Findings

Achieves superior performance on observability datasets.

Maintains similar accuracy on standard benchmarks.

Enables more accurate long-term predictions.

Abstract

We introduce the Cisco Time Series Model, a univariate zero-shot forecaster. This time series foundation model is the result of a general architectural innovation to a time series model enabling it to accept multiresolution input, applied to a popular decoder-only time series model (TimesFM). The resulting multiresolution decoder-only model is trained on over 300B unique data points, with more than half coming from the observability domain. Quantitative and qualitative evaluations demonstrate that the resulting model achieves superior performance on observability datasets while retaining very similar performance on a standard general-purpose forecasting benchmark (GIFT-Eval), and suggest that the multiresolution structure enables the model to make more accurate predictions on long context input.