Treatment of Missing Values for Association Rules: A Recent Survey

Abstract

Missing values and incomplete data are usual occurrences in real datasets [1]. The problem of recovering missing values from a dataset has become an important research issue in the field of data mining and machine learning [2]. With the speedy increase in the use of databases, the difficulty of missing values unavoidably arises. The techniques developed to effectively recover these missing values should be highly accurate in order to remove the missing values completely. The association rules are the popular method that is effectively used to establish the relationship among items in databases. The discovered association rules are useful to recover the missing values in databases. There are several methods proposed to surmount the problem of missing value [4]. In this paper we present a study over the existing methods.

Enhanced Robust Association Rules (ERAR)Method for Missing Values Imputation

“…Knowing the missing values mechanisms and patterns can offer performance advantage in missing values imputation algorithms. Nevertheless, it is difficult to find a method that works best in all cases [1]. For this reason, finding missing values imputation methods that can offer optimal performance for specific cases is an open problem.…”

“…Missing values is a common problem highlighted in data mining, statistics, and database domains [23][24][25]. Mining techniques are typically unable to deal with missing values and need some kind of pre-processing or workaround [1]. Imputation is a technique used to handle missing values where missing values are replaced by substituted values [26].…”

“…Based on their method, Ragel and Crmilleux (1998) developed the Robust Association Rules (RAR) method to discover missing values [7]. Since then, more methods for handling missing values were proposed based on the RAR approach, such as Recycle Composed Association Rule (RCAR) [8], Fast Recycle Combined Association Rules (FRCAR) [1], iterative missing-value completion [9] and Association Rule Mining from Data with Missing Values (ARDM) [10].…”