Analysis of bee hive data using robust methods

Abstract

Bees are of utmost importance for food production and agriculture, since without them pollination of many species of plants would not be possible. Keeping bees in good health is thus not only of biological but also of economic importance. New technologies allow to measure weights of bee hives with high frequency. These signals depend on many factors such as wind speed, air pressure, temperature and humidity. Data sets need to be pre-processed to adequately deal with outliers and interventions from bee keepers. After this data cleaning step using robust techniques, changes in hive weights and covariates are observed and should give an indication of the health status of the bees. Comparison of multiple hive weight data of one location reveal irregular behaviour of a single bee hive in relation to the others. For this aim, Regression Methods and Principal Component Analysis are used in a specially adapted manner.