Misalignment problem in matrix decomposition with missing values

Data collection within a real-world environment may be compromised by several factors such as data-logger malfunctions and communication errors, during which no data is collected. As consequence, appropriate tools are required to handle the missing values when analysing processing data. This problem often tackled via matrix decomposition. While it has been successfully applied in wide range of applications, this work we report an issue that neglected literature “degenerates” quality imputations obtained decomposition multivariate time-series (with smooth evolution). Briefly, consists misalignment result: fall incorrect transitions between observed imputed not smooth. We address proposing post-processing alignment strategy. According our experiments, adjustment substantially improves accuracy (when occurs). Moreover, results also suggest occurs mostly dealing with small number due lack generalization ability.