The inefficient handling of analysis and the negative impact to model performance due to missing values is a common problem within the time series data. This research presents MEOW (Mapping Error Onto Weight), a new approach for time-series data imputation. It adopts a static ensemble approach by employing an adaptive weighting scheme that features both forward and backward projected values. MEOW differs greatly from static ensemble methods because it employs linearly changing weights throughout the entire inferred series, which helps mitigate the issue of being far behind or beyond the predicted values. Thus, this approach improves the estimated values by reducing the overall inference error. Through rigorous experiments using real-world datasets of widely differing characteristics, MEOW was shown to outperform all other known methods and provide the lowest inference error and highest robustness in the greatest number of circumstances.