WalkForward

PortfolioOptimisers.IndexWalkForward Type

struct IndexWalkForward{T1, T2, T3, T4, T5} <: WalkForwardEstimator
    train_size
    test_size
    purged_size
    expend_train
    reduce_test
end

Implements index-based walk-forward cross-validation for time series, supporting purging and flexible train/test windowing.

Fields

• train_size: Number of observations in each training set.

• test_size: Number of observations in each test set.

• purged_size: Number of observations to exclude from the end of each train set adjacent to a test set.

• expend_train: If true, training set always starts from the beginning; otherwise, uses a rolling window.

• reduce_test: If true, reduces the last test set to fit remaining data; otherwise, drops incomplete test sets.

Constructors

IndexWalkForward(train_size::Integer, test_size::Integer; purged_size::Integer = 0,
                 expend_train::Bool = false, reduce_test::Bool = false)

Keyword arguments correspond to the fields above.

Validation

• train_size, test_size, and purged_size must be non-empty, non-negative, and finite.

• Ensures train_size + purged_size < T where T is the total number of observations.

Examples

julia> IndexWalkForward(100, 20; purged_size = 5, expend_train = true, reduce_test = false)
IndexWalkForward
    train_size ┼ Int64: 100
     test_size ┼ Int64: 20
   purged_size ┼ Int64: 5
  expend_train ┼ Bool: true
   reduce_test ┴ Bool: false

Related

• [WalkForwardEstimator]-(@ref)

• [WalkForwardResult]-(@ref)

• [split]-(@ref)

• [n_splits]-(@ref)

source

PortfolioOptimisers.DateWalkForward Type

struct DateWalkForward{T1, T2, T3, T4, T5, T6, T7, T8, T9} <: WalkForwardEstimator
    train_size::T1
    test_size::T2
    period::T3
    period_offset::T4
    purged_size::T5
    adjuster::T6
    previous::T7
    expend_train::T8
    reduce_test::T9
end

Implements date-based walk-forward cross-validation for time series, supporting flexible windowing, purging, and custom date adjustment.

Fields

• train_size: Number of periods or length of training window (integer or date period).

• test_size: Number of observations in each test set.

• period: Step size for each split.

• period_offset: Optional offset to shift the split dates.

• purged_size: Number of observations to exclude from the end of each train set adjacent to a test set.

• adjuster: Function or estimator to adjust the date range (e.g., for business days).

• previous: If true, allows test indices to use previous available date if exact match not found, else use the next available date.

• expend_train: If true, training set always starts from the beginning; otherwise, uses a rolling window.

• reduce_test: If true, reduces the last test set to fit remaining data; otherwise, drops incomplete test sets.

Constructors

DateWalkForward(train_size::IntPeriodDateRange, test_size::Integer;
                period::DatesUnionPeriod = Dates.Day(1),
                period_offset::Option{<:DatesUnionPeriod} = nothing,
                purged_size::Integer = 0, adjuster::DateAdjType = identity,
                previous::Bool = false, expend_train::Bool = false,
                reduce_test::Bool = false)

Keyword arguments correspond to the fields above.

Validation

• test_size and purged_size must be non-empty, non-negative, and finite.

• If train_size is an integer, it must be non-empty, non-negative, and finite.

Examples

julia> DateWalkForward(252, 21; period = Dates.Day(1), purged_size = 5, expend_train = true)
DateWalkForward
     train_size ┼ Int64: 252
      test_size ┼ Int64: 21
         period ┼ Dates.Day: Dates.Day(1)
  period_offset ┼ nothing
    purged_size ┼ Int64: 5
       adjuster ┼ typeof(identity): identity
       previous ┼ Bool: false
   expend_train ┼ Bool: true
    reduce_test ┴ Bool: false

Related

• [WalkForwardEstimator]-(@ref)

• [WalkForwardResult]-(@ref)

• [split]-(@ref)

• [n_splits]-(@ref)

source