A large language model is a foundation model for natural-language processing, trained to perform tasks such as question answering, text classification, translation, and text generation. Its core capability is language generation rather than time-series forecasting. section 8.10 section 1.4
For forecasting, convert a numerical sequence into a natural-language prompt, ask for the next values, and parse the generated string back into predictions. This approach treats forecasting as next-token-like prediction, but requires careful prompt construction and output processing. section 8.0 section 8.10
Practical considerations shape whether an LLM is appropriate for forecasting:
- Use prompting when a natural-language interface is required and robust prompts can be constructed. section 8.0
- Experiment with zero-shot, few-shot, and chain-of-thought prompts; few-shot and chain-of-thought prompting usually produce the best results. section 8.10 section 8.9
- Expect extra preprocessing and extraction code, substantial computational requirements for local deployment, and potentially poor forecasting performance. section 8.10 section 8.9
Prompt-based forecasting
Flan-T5 and Llama-3.2-3b-instruct can receive programmatically constructed prompts containing an input series and return a string whose values are extracted as forecasts. They can also be used locally or through API calls. section 8.0 section 8.10
Table 8.3 Pros and cons of each foundation forecasting model
| Model |
Pros |
Cons |
When to use |
|---|---|---|---|
| TimeGPT |
Easy and fast to use Comes with many native functions Works on any device, regardless of hardware Free plan |
Paid product for a certain usage Model may not be available if the server is down. |
Forecasting on long and short horizons with exogenous features. You need to fine-tune but do not have the local resources. |
| Lag-Llama |
Open source model Free to use |
Awkward to use because we must clone the repository Speed of inference depends on our hardware. |
Quick proof of concept Ideal for research-oriented projects
|
| Chronos |
Open source model Free to use Can be installed as a Python package |
Speed and accuracy depend on our hardware. Fine-tuning requires cloning the repository. We must define some functions manually. |
Forecasting on horizons shorter than 64 time steps Forecasting series without strong trends |
| TimesFM |
Open source model Free to use Can be installed as a Python package Supports exogenous features |
Only the largest model is available. Restrictive requirements for use (16 GB of RAM, Linux-based system to install dependencies) Knowledge of JAX is required for fine-tuning. Cannot perform anomaly detection because it is a deterministic model |
Ideal for deterministic forecasting in which prediction intervals are not required You want to feed a long input series. |
| Moirai |
Open source model Free to use Supports exogenous features Can be installed as a Python package |
Speed and accuracy depend on the hardware. We must define helper functions when not working entirely with GluonTS. |
You have exogenous features. Forecasting on horizons shorter than 256 time steps |
| Large language models (Flan-T5 and Llama-3.2) |
Possible to use free LLMs Interacting through natural language feels intuitive, |
Not built to handle time-series data Requires lengthy code to preprocess the input and extract the values from the output Requires a lot of computation power for local setups |
You already have access to an LLM. You need a natural language interface for forecasting. You know how to develop prompts and safeguards for edge cases. |
Compare LLM-based forecasting with purpose-built foundation forecasting models by hardware needs, deployment, speed, and suitable use cases. section 8.9
