MultiStream supports a JSON file containing time series organized in a hierarchy structure. We provide a script that helps you generate this file. Follow the instructions below to create a compatible dataset and load it into MultiStream.

Preparing your data

To start, you need two input files:

1. CSV File: Time Series Data
   This file must contain the following structure: (example)
   • First column: it must be named date and formatted using ISO-8601, either as YYYY-MM-DD or YYYY-MM-DD HH:mm:ssZ
     • The Z suffix is required if using UTC
     • Suported time granularities: minutes, hours, days, weeks, months and years
     • Examples: 2000-01-01 or 2016-05-28 19:50:00
   • Other columns: time series category names. These must be strings identifying each individual time series
   
2. JSON File: Hierarchical Structure
   This file should define the hierarchy of your categories (example)

Generating a supported file

We provide a preprocessing script that transforms your CSV and JSON file into a format that MultiStream can load.

Prerequisites

• Node.js version 14 or later
• Download the required script and dependencies: generate_supported_file.zip
• Open your terminal (Linux/macOS) or command prompt (Windows), navigate to the downloaded folder, and run: npm i

Run the script

Use the command line to execute preprocessing.js with the following arguments:

Argument	Description
--raw (required)	Path to the CSV file with the raw time series data
--hierarchy (required)	Path to the JSON file containing the hierarchy
--output (required)	Path to save the output JSON file
--granularity (required)	Time unit used in the series Allowed values: minutes, hours, days, weeks, months, years
--step (required)	Step size between time points (usually 1)
--datatype (optional)	Description of the time series data (e.g., "number of bands")

You can run the script using this command:

node preprocessing.js --raw=input/data.csv --hierarchy=input/hierarchy.json --output=output.json --granularity=years --step=1 --datatype="number of bands"
							

This will generate a file named output.json, which can be loaded by MultiStream.

Loading the file in MultiStream

Go to the MultiStream web application and upload the generated file (e.g., output.json).

Tips & requirements

CSV File: Time Series Data

• Use clear, unique, and non-empty column names.
• The first column must be named date and follow the ISO-8601 format (e.g., YYYY-MM-DD or YYYY-MM-DD HH:mm:ssZ).
• Avoid missing or null values in any column.

JSON File: Hierarchical Structure

• Define a clear hierarchical structure for the time series categories.
• Ensure every category in the hierarchy matches a column name in the CSV (excluding date).
• Use valid JSON syntax and ensure correct nesting of hierarchy nodes.

If you have any questions or run into issues, feel free to contact us