Clustering and Association Rules Based Missing Data Filling Algorithm

Filling missing data is crucial to preprocessing data. Missing data reduces the accuracy of statistics, causes statistical errors. Traditional methods for dealing with missing data include simple deletion, mean filling, or simply ignoring missing values. However, these methods have limitations in accuracy and computational feasibility. Therefore, it is essential to select a effective method to fill missing data. In this paper, we propose a method based on Fuzzy c-means clustering algorithm, Dynamic division of connected graph algorithm and Association Rules named FDAR to fill missing data. First, we use Fuzzy c-means clustering (FCM) and Dynamic division of connected graph (CABDDCG) algorithm to cluster, and then mine association rules (AR) in each part of cluster. Next, if there are association rules, the association rules are filled; Otherwise, the nearest first
$k$
values in the same cluster are used to fill. Experiments show FDAR algorithm in filling missing data has high accuracy on six UCI datasets.