ForecastBot

ForecastBot is a web application designed to provide users with an intelligent chatbot interface that helps them select and interact with forecasting models, powered by a Large Language Model (LLM). The chatbot guides users through the process of selecting a model based on their dataset and specific parameters, using Streamlit for the user interface and FastAPI for the backend.

The app integrates both advanced transformer-based forecasting models (such as Informer, PatchTST, Pyraformer) and traditional models (ARIMA, SARIMAX), offering a seamless experience for model selection, dataset configuration, and result visualization.

Architecture

Below is the architectural pattern of ForecastBot, showcasing the flow between its main components: the client, the server, and the forecasting models.

Features

• Chatbot Interface: An interactive chatbot built with LLM capabilities, allowing users to interact naturally and get model recommendations based on their input.
• Model Support: Includes the following forecasting models:
  • Informer
  • PatchTST
  • Pyraformer
  • ARIMA
  • SARIMAX
• Visualization: Interactive graphs and charts to display model performance.

Default Dataset

By default, the models use the ETTh1 dataset, which has been tested by training and evaluating models using both GPU and CPU environments. The GPU testing was conducted using a container built with RapidsAI, leveraging the power of CUDA for accelerated computation. Similarly, the models were also tested on a CPU-only environment for comparison.

You can download the dataset from the following link: ETTh1.csv.

Ensure the dataset is placed in the following directory structure:

ForecastBot/
└── server/
  └── data/
    └── ETTh1.csv

Local environment

How to start (both on Linux and Windows (PoweShell/WSL)):

$: ./forecastbot dev up --build

or any parameter of docker compose you prefer after dev.

Pretrain models using dev (inside server container)

$: python /server/src/transformer/Arima/test.py     # Pretrain Arima Model 
$: python /server/src/transformer/Sarimax/test.py   # Pretrain Sarimax Model  
$: python /server/src/transformer/PatchTST/test.py  # Pretrain PatchTST Model

Prod configuration

$: ./forecastbot prod up -d --build