FUSAIN: Combining Functional Dependencies and Clustering for Missing Values Imputation

Missing data is a common problem faced with real-world datasets. A large number of missing data will greatly affect the quality of the data and cause deviations in the results of data analysis. Therefore, missing values imputation (MVI) is a critical data processing process. Most imputation methods model the distribution of observed data to approximate the missing values. Such an approach usually models a single distribution for the entire dataset, which ignores the dependencies between data. In this paper, we propose a novel hybrid imputation algorithm, called combining Functional dependencies and clUstering for miSsing vAlues ImputatioN (FUSAIN), which combines Functional Dependencies (FDs), K Nearest Neighbor (KNN), and Affinity Propagation (AP) clustering. This proposed algorithm not only considers the distribution of data but also uses the data dependency relationship represented by FDs to impute missing values. From the experimental results, the imputation performance of the proposed algorithm achieves superior performance compared to common and popular imputation algorithms.