Cluster-based trajectory segmentation with local noise

An Uncertainty-based Neural Network for Explainable Trajectory Segmentation

As a variant task of time-series segmentation, trajectory segmentation is a key task in the applications of transportation pattern recognition and traffic analysis. However, segmenting trajectory is faced with challenges of implicit patterns and sparse results. Although deep neural networks have tremendous advantages in terms of high-level feature learning performance, deploying as a blackbox seriously limits the real-world applications. Providing explainable segmentations has significance for result evaluation and decision making. Thus, in this article, we address trajectory segmentation by proposing a Bayesian Encoder-Decoder Network (BED-Net) to provide accurate detection with explainability and references for the following active-learning procedures. BED-Net consists of a segmentation module based on Monte Carlo dropout and an explanation module based on uncertainty learning that provides results evaluation and visualization. Experimental results on both benchmark and real-world datasets indicate that BED-Net outperforms the rival methods and offers excellent explainability in the applications of trajectory segmentation.

Analysis of movement recursions to detect reproductive events and estimate their fate in central place foragers

BACKGROUND

Recursive movement patterns have been used to detect behavioral structure within individual movement trajectories in the context of foraging ecology, home-ranging behavior, and predator avoidance. Some animals exhibit movement recursions to locations that are tied to reproductive functions, including nests and dens; while existing literature recognizes that, no method is currently available to explicitly target different types of revisited locations. Moreover, the temporal persistence of recursive movements to a breeding location can carry information regarding the fate of breeding attempts, but it has never been used as a metric to quantify recursive movement patterns. Here, we introduce a method to locate breeding attempts and estimate their fate from GPS-tracking data of central place foragers. We tested the performance of our method in three bird species differing in breeding ecology (wood stork (Mycteria americana), lesser kestrel (Falco naumanni), Mediterranean gull (Ichthyaetus melanocephalus)) and implemented it in the R package 'nestR'.

METHODS

We identified breeding sites based on the analysis of recursive movements within individual tracks. Using trajectories with known breeding attempts, we estimated a set of species-specific criteria for the identification of nest sites, which we further validated using non-reproductive individuals as controls. We then estimated individual nest survival as a binary measure of reproductive fate (success, corresponding to fledging of at least one chick, or failure) from nest-site revisitation histories during breeding attempts, using a Bayesian hierarchical modeling approach that accounted for temporally variable revisitation patterns, probability of visit detection, and missing data.

RESULTS

Across the three species, positive predictive value of the nest-site detection algorithm varied between 87 and 100% and sensitivity between 88 and 92%, and we correctly estimated the fate of 86-100% breeding attempts.

CONCLUSIONS

By providing a method to formally distinguish among revisited locations that serve different ecological functions and introducing a probabilistic framework to quantify temporal persistence of movement recursions, we demonstrated how the analysis of recursive movement patterns can be applied to estimate reproduction in central place foragers. Beyond avian species, the principles of our method can be applied to other central place foraging breeders such as denning mammals. Our method estimates a component of individual fitness from movement data and will help bridge the gap between movement behavior, environmental factors, and their fitness consequences.