Visualization of multiple influences on flight behavior by the Viscovery® data mining tool
G. Kastberger1 and G. Kranner2
1 Institute of Zoology, University of
Graz, Graz, Austria
2 Eudaptics, Vienna, Austria
 
Viscovery® SOMine (Self Organizing Maps) is a powerful tool for exploratory data analysis and data mining. Employing Kohonen neural networks, it puts complex data into order based on its similarity, and shows a map from which the features of the data can be identified and evaluated. This tool is very fast on large data sets, through the implementation of new techniques (such as SOM scaling) the speed in creating maps is notably increased compared to the original SOM algorithm. Designed also for scientific research, it supports analysis of non-linear dependencies, parameter-free clustering, association and prediction, non-linear regression pattern recognition and animated system states monitoring.
With this data mining tool we mapped non-linear relationships in multi-dimensional behavioral data assessed from honeybees under free-flight regime: Apis mellifera in Graz [3] and Apis dorsata in New Delhi [2]. In the experiments we trained fully sighted and ocelli-occluded bees to come to an artificial feeder. On the way from the nest to the feeder (feeding flight) and from the feeder to the nest (homing flight), respectively, the bees had to pass a flight chamber with a stationary side light (which was on or off). Hereby, we considered the following flight parameters: the direction of the bees long axis (L), the flight course (C), the deviance of L and C from the target direction (Lt, Ct), the angular velocities of L and C (dL, dC), the deviance of L from C (Lc = yawing), ground velocity (Vg) and acceleration (Ag). The behavioral responses were mapped by Viscovery SOMine (a) using the original data and (b) using contracted data processed from the original ones by conventional statistics (arithmetical means and standard deviations) and by case statistics (considering probabilities of positively or negatively oriented responses, respectively [1]).
For both honeybee species it is proved that occlusion of ocelli will not cut down orientational abilities, it (significantly) enhances or reduces reactivity under specific surrounding conditions (side light stimulation, spatial position in the flight chamber). However, the differences in behavior found between feeding and homing condition, between side light on or off, regarding the spatial influences of length or width position in the flight chamber are more dominant than those between trained fully sighted and ocelli-occluded honeybees.
© 1998 Noldus Information Technology b.v.