synthetic-sample

synthetic_sample

synthetic_sample is a data generation application for producing synthetic sales transactions over a time series, including associated shipment and product data

Usage

Sample data is generated by running synthetic_sample_generator.py and using

python3 synthetic_sample_generator.py --json_filepath JSON_FILEPATH --output_directory OUTPUT_DIRECTORY --create_records

where

• json_filepath is the filepath to the input JSON (see Request Requirements below)
• output_directory is the directory to save output data to, in CSV format
• create_records is a flag that indicates that raw record data should also be saved to the output directory. Running without this flag results in only aggregate output data

Request Requirements

The required input format is a JSON with the following fields:

• Required:
  • start_date: date in the first period to include, e.g. if 2020/02/15 is provided, the full week of that date will be included
  • end_date: date in the last period to include, e.g. if 2020/02/15 is provided, the full week of that date will be included
  • annual_growth_factor: year over year growth factor, 10% growth corresponds to a value of 1.1
  • period_type: indicates what type of curve to generate, supports "month" or "week"
  • at least one of
    • total_sales: total number of sales for the period
    • total_packages: total number of packages shipped for the period
    • total_quantity: total number of items sold for the period
    • annual_sales: annualized number of sales for the period
    • annual_packages: annualized number of packages shipped for the period
    • annual_quantity: annualized number of items sold for the period
  • curve_definition: Definition of the curve to create, either as a list of dictionaries with each feature or as a string indicating the name of the default curve to use.
    • If a list of dictionaries is provided, they must adhere to the following structure
      • Required Keys:
        • anchor_type: Type of annual anchor used to define the feature
          • Possible Values: "holiday", "week_of_year", "month_of_year", "day_of_year"
        • anchor_point: Annual point to define the feature
          • Possible values: (string) - holiday name, (int) - week or day of year
        • anchor_value: Cumulative percent of total sales (0.0-1.0) completed by the end of the period of the anchor_point
      • Optional Keys:
        • relative_start: Number of periods before the anchor_point to define a relative cumulative percent value
        • start_value: Cumulative percent of total sales (0.0-1.0) completed by the end of the period indicated by relative_start
        • relative_end: Number of periods before the anchor_point to define a relative cumulative percent value
        • end_value: Cumulative percent of total sales (0.0-1.0) completed by the end of the period indicated by relative_end
    • If a string is provided, it must correspond to a default in synthetic_sample/defaults/curves/{period_type}/{curve_definition}.json
      • Initial set of available curves are
        • modern_brand
        • modern_distributor
        • traditional_brand
        • traditional_distributor
• Optional:
  • default_type: string indicating the type of defaults to use, these can be found as JSON in synthetic_sample/defaults/lib/
  • product_distribution: dictionary of product labels (i.e. SKUs) and their relative weights
  • week_distribution: dictionary of weeks of the month (where 1 is the first week and -1 is the last) and their relative weights
  • weekday_distribution: dictionary of weekdays (where 0 is Monday and 6 is Sunday) and their relative weights
  • seasonal_distribution: dictionary of seasons ("Q1"..."Q4") and their relative weights
  • modifiers: list of any modifiers to apply.
    • "covid": Applies a 33% boost to all periods between 2020/3/26 and 2021/9/1

Example:

The below request will generate data for each month starting 2018-06 and ending 2020-12.

{
  "start_date": "2018-06-01",
  "end_date": "2020-12-31",
  "total_sales": 1000000,
  "total_packages": 1500000,
  "total_quantity": 6000000,
  "annual_growth_factor": 1.15,
  "product_distribution": {
    "AAA-01" : 1,
    "AAA-02" : 2.5,
    "AAA-11" : 5.6,
    "BBB-10" : 0.5,
    "BBB-20" : 1
  },
  "week_distribution": {
    "1": 0.1,
    "-1": 0.5
  },
  "weekday_distribution": {
    "0": 0.0,
    "1": 0.0,
    "2": 0.0,
    "3": 0.0,
    "4": 0.0,
    "5": 2.0,
    "6": 1.0
  },
  "seasonal_distribution": {
    "Q1": 1,
    "Q2": 1,
    "Q3": 1,
    "Q4": 1
  },
  "period_type": "month",
  "curve_definition": [
    {
      "anchor_type": "month_of_year",
      "anchor_point": 1,
      "anchor_value": 0.0424
    },
    {
      "anchor_type": "month_of_year",
      "anchor_point": 2,
      "anchor_value": 0.103
    },
    {
      "anchor_type": "month_of_year",
      "anchor_point": 3,
      "anchor_value": 0.203
    },
    {
      "anchor_type": "month_of_year",
      "anchor_point": 4,
      "anchor_value": 0.3152
    },
    {
      "anchor_type": "month_of_year",
      "anchor_point": 5,
      "anchor_value": 0.4139
    },
    {
      "anchor_type": "month_of_year",
      "anchor_point": 6,
      "anchor_value": 0.4776
    },
    {
      "anchor_type": "month_of_year",
      "anchor_point": 7,
      "anchor_value": 0.5321
    },
    {
      "anchor_type": "month_of_year",
      "anchor_point": 8,
      "anchor_value": 0.5897
    },
    {
      "anchor_type": "month_of_year",
      "anchor_point": 9,
      "anchor_value": 0.6715
    },
    {
      "anchor_type": "month_of_year",
      "anchor_point": 10,
      "anchor_value": 0.7836
    },
    {
      "anchor_type": "month_of_year",
      "anchor_point": 11,
      "anchor_value": 0.9018
    },
    {
      "anchor_type": "month_of_year",
      "anchor_point": 12,
      "anchor_value": 1.0
    }
  ]
}