Increasing Tiny Data Imputation Accuracy Using Temporal Polynomial Interpolation

Missing data is a common case in many systems. Incomplete data can affect the accuracy of analysis results. There are several methods have been proposed to handle missing data. Commonly, the way to recover the data is by eliminating the incomplete data series or replacing it with the average value of existing or previous data. This paper proposes a more accurate imputing incomplete data using the temporal polynomial interpolation method for tiny data such as live data. The result shows quadratic interpolation method generally gives fairly accurate results compared to other polynomial interpolation methods, with an MSE value of 0.051 for 10% missing data, 0.033 for 30% missing data, and 0,035 for 50% missing data. This MSE result is still better than another method like simple regression, which has an MSE value for 10% missing data around 0.71, 30% missing data around 0.59, and 50% missing data around 0.44.