atspR: An R package for automated time series preprocessing
Sueppong Mueanchamnong, Pattharaporn Thongnim · SoftwareX · 2026
Preprocessing sensor-derived time series data is a critical but error-prone step in applied machine learning, frequently impeded by temporal gaps, mixed-type encodings, silent missing-value handling, and data leakage introduced during feature scaling and dataset splitting. This paper presents atspR , an open-source R package designed for automated, leakage-free preprocessing of time series and tabular data across scientific domains including agriculture, environmental monitoring, marine science, and hydrology. The package orchestrates ten sequential stages, timestamp repair, missing-value standardization, type coercion, chronological splitting, imputation, visualization, leakage-free feature scaling, and walk-forward cross-validation through a single function call, while exporting each stage independently for flexible custom use. Plain-language console output explains every automated decision, making the pipeline accessible to researchers and domain practitioners regardless of programming background.