Abstract

This invention introduces a novel data imputation algorithm named as Seasonal Mean Imputation (SMI), designed to address the challenge of dealing with missing values in data preprocessing stage for tasks related to data science or Machine Learning (ML) implementation. In contrast to conventional Mean Imputation (MI) technique that involves filling of all the missing values in a dataset with only mean of the whole data, this novel approach improves upon by seasonally imputing or filling the missing values pertinent to the seasonality of the data such that original data’s seasonality pattern is considered. The seasonality is a mandatory part of all the ML  implementations because ML algorithms’ core purpose is to learn the patterns in data and predict future data according to the past data patterns. The goal of this algorithm is to improve the overall predictive accuracy of the ML models with a similar complexity cost as incurred by the traditional MI technique.

Keywords: Seasonal mean imputation, Mean imputation, Seasonal mean imputation versus mean imputation, Data imputation techniques, Seasonal mean, Data filling, SMI