Artificially introduced MNAR dataset for missing data analysis in activity recognition

This dataset includes Missing Not at Random (MNAR) data artificially introduced into activity recognition measurements collected from a smart bracelet. The dataset is tailored for research on the challenges of imputing non-randomly missing data in wearable sensor datasets. Original data were recorded during walking, running, climbing stairs, and resting activities, and missing values have been systematically introduced under MNAR conditions. This dataset is ideal for testing imputation algorithms that address complex missing data patterns affecting data quality and activity classification accuracy.