GARDA: Granular Association-rule-based Data Imputation Approach for IoT Sensor Networks

Data imputation is the process of replacing missing data with substituted values to maintain the integrity and usability of a dataset. It is critically important for improving the analysis of data collected from Internet of Things (IoT) sensors and devices, which are often affected by failures or communication issues that can result in data loss. The absence of missing data in IoT applications is of paramount relevance to avoid errors, accidents, and wrong decisions. This work introduces GARDA-Granular Association-rule-based Data imputation Approach-a data imputation method based on association rules and granular computing for imputing missing data coming from IoT sensors networks characterized by sensors data with high correlation. GARDA performs the granulation of continuous data with interval-valued sets from which granular association rules are mined. These association rules are used to impute the missing intervals. Finally, the intervals are transformed back to continuous data to impute missing values. The approach has been evaluated on the Intel Lab and Beijing Air datasets and compared with state-of-the-art techniques, demonstrating good performance.