PyPOTS

Why we included this project

Real-world sensor and industrial data rarely arrive as clean, regularly sampled tables. Missing readings and irregular timestamps are the norm, and PyPOTS is a PyTorch toolbox built around exactly that reality. It bundles 50+ neural network models for imputation, classification, clustering, forecasting, and anomaly detection on multivariate time series with NaN gaps, so instead of hand-rolling masking strategies or juggling incompatible model repos, you get one consistent API. Familiar architectures like Transformer and iTransformer are adapted to accept partially observed inputs directly, which helps teams in healthcare, manufacturing, energy, or hydrology where incomplete records are routine. It is well documented with tutorials, making it a practical starting point for engineers prototyping pipelines and researchers benchmarking methods.