ts-data-generator 0.2.0

A Python library for generating synthetic time series data

Synthetic Time Series Data Generator

A Python library for generating synthetic time series data

^{Special thanks to: Nike-Inc repo}

Installation

PyPi (recommended)

You can install with pip directly by

pip install ts-data-generator

Repo

After cloning this repo and creating a virtual environment, run the following command:

pip install --editable .

Usage

1. To check out constructing for time series data, check the sample notebook here
2. To extract the trends from an existing data, check this sample notebook here

CLI

You can also use the command line utility tsdata to generate the data.

(venv) ~/ts-data-generator   cli $ tsdata generate --help
Usage: tsdata generate [OPTIONS]

  Generate time series data and save it to a CSV file.

Options:
  --start TEXT                    Start datetime 'YYYY-MM-DD'  [required]
  --end TEXT                      End datetime 'YYYY-MM-DD'  [required]
  --granularity [FIVE_MIN|HOURLY|DAILY]
                                  Granularity of the time series data  [required]
  --dims TEXT                     + separated list of dimensions definition of format 'name:function:values'  [required]
  --mets TEXT                     + separated list of metrics definition trends of format 'name:trend(*params)'  [required]
  --output TEXT                   Output file name  [required]
  --help                          Show this message and exit.

For example you can call this cli tool like below to generate data

tsdata generate \
  --start "2019-01-01" \
  --end "2019-01-12" \
  --granularity "FIVE_MIN" \
  --dims "product:random_choice:A,B,C,D" \
  --dims "product_id:random_float:1,4" \
  --dims "const:constant:5" \
  --mets "sales:LinearTrend(limit=500)+WeekendTrend(weekend_effect=50)" \
  --mets "trend:LinearTrend(limit=10)" \
  --output "data.csv"

Release method

1. git tag <x.x.x>
2. git push origin <x.x.x>