DataSplits.jl: Data splitting strategies for machine learning in Julia

DataSplits.jl is a Julia package (Bezanson et al., 2017) for constructing training, validation, test, and cross-validation partitions in scientific machine learning workflows.While random train/test splits and standard k-fold cross-validation are widely available in machine learning libraries, many scientific applications require more controlled splitting procedures.The package exposes a single entry point: res = partition(data, strategy; kwargs...) Supported strategies include distance-based representative sample selection for calibration modelling, group-aware validation to avoid leakage between related observations, and timeaware validation for temporally ordered data.The package is lightweight and interoperable with MLUtils.jl,Flux.jl, and MLJ.jl.It can be installed from the Julia General Registry with ] add DataSplits.