Binary Constraint

Overview

Binary Constraints in Verteego enable direct comparison or interaction between two columns within a dataset, which can be either variable or static. This type of constraint is applied on a row-by-row basis, making it suitable for conditions where variables within the same dataset must maintain a specific relationship to one another.

Application

Binary Constraints are commonly used to enforce relationships such as ensuring one value remains greater than or less than another (e.g., a product’s price must always be higher than a competitor’s price).

Parameters

• where: An optional dictionary of key-value pairs that filters rows based on specified conditions. For instance, setting where={"month": [1, 2, 3], "year": 2022} would apply the constraint only to rows from the first quarter of 2022, ignoring all others.

• left_column: Specifies the column containing the variable or static number to be constrained.

• left_column_weight: Optional. A column containing weights to be applied to values in the left_column.

• operator: Defines the relationship between the left and right column values. Options are equal, lesser (less than or equal to), and greater (greater than or equal to).

• right_column: Specifies the column against which the left_column is being constrained.

• right_column_weight: Similar to left_column_weight, for applying weights to the right_column.

• delta: An optional numerical value that offsets the constraint, adding flexibility to the condition (e.g., price must be greater than competition_price + 1).

• relax: Indicates whether the constraint can be relaxed (with a penalty that gets reported). Defaults to false.

Examples

Below are examples of YAML configurations for binary constraints:

• Setting a Minimum Price Above Competition:

yamlCopy codecompetition_lower_bound:
  constraint_type: binary
  left_column: price
  operator: greater
  right_column: competition_price
  delta: 0
  relax: false

• Setting a Maximum Price Below an Upper Limit:

yamlCopy codecompetition_upper_bound:
  constraint_type: binary
  left_column: price
  operator: lesser
  right_column: competition_upper_bound_price
  delta: 0
  relax: false

Best Practices

When implementing Binary Constraints, it is crucial to:

• Select relevant columns for the left and right sides of the equation to ensure meaningful comparisons.

• Use appropriate weights and deltas to refine the conditions under which constraints apply.

• Consider using the relax parameter for scenarios where strict adherence to constraints might not always be possible or desirable, allowing for more flexible optimization solutions.