NSF funded spatio-temporal data mining research enables behavioral ecologists to understandthe relationship between dominance rank andfemale space usage among chimpanzees [1]

Since 1960, researchers (led byDr.JaneGoodall) have been collecting and analyzing data about the social behavior of chimpanzees (Pan troglodytes) at Gombe National Park, Tanzania. The preservation of one of the largest and richest animal behavior datasets has motivated its digitization, which is now housed at the University of Minnesota. However, manual analysis of this data remains a time consuming and difficult task. These challenges motivated an inter-disciplinary collaboration between computer scientists and behavioral ecologists at University of Minnesotato develop spatio-temporal data miningtechniques that enablemore efficient analyses of this dataset and others.

Since chimpanzees are our closest living relatives, studying them is important for understanding human evolution. Chimpanzees are a highly gregarious species with transient fission-fusion groupings among individuals in the community. Sociability and ranging patterns alsovary among differentchimpanzee populations. Understanding the factors that influence this variability is of primary interest to behavioral ecologists. Also, understanding female space use is particularly crucial because it is thought to determine male distribution, inter-group aggression, and mating systems. Our research aimed to investigate female space use among the chimpanzees of GombeNational Park. Figure 1illustrates the distribution of spatial locations where females were observed during 2001-2002.

While researchers suspected that dominance rankmay influenceindividual space usage among females, the exact nature of the relationship was not well-understood. This project therefore focused on developing novel spatio-temporal data mining techniques to investigate this relationship[1,5]. Results demonstrated that high-ranking females had smaller core areas and that size differences were particularly pronounced during periods of food scarcity. Immigrants used areas further away from dominant females, and subordinates had lower site fidelity [2]. These patterns suggested that dominant females out-compete subordinates, forcing them to settle elsewhere, range more widely, and shift their space use across time. These findings improved our understanding of chimpanzee space use and the overall social structure of this species.We are pursuing further research through which we will apply these techniques to male ranging patterns [2]. While contributing greatly to understanding chimpanzee behavior, these findings are also applicable to the broader behavioral ecology domain and our techniques could be used in other observational datasets.

The use of classical data mining techniques, such as dendrograms to group females provided ecologists an intuitive, easy-to-understandmeans ofidentifying the presence of two neighborhoods of females, namely northern and southern.However, the limitation of these approaches motivated Computer Science research, resulting in new spatio-temporal data mining methods to clusterspatial point processes [3], and spatial co-location mining [4]. The new spatio-temporal data mining techniques developed have applicability to any dataset that has spatial characteristics, including transportation, public safety, public health, troop movements, etc.

References

  1. Murray, C.M. and Mane, S. (2005). Neighborhood identification in female chimpanzees (Pan troglodytes) using spatio-temporal clustering. Poster presented at the Annual Meeting of the Animal Behavior Society,Utah.
  2. CarsonMurray (2006). The Influence of Food Competition on Foraging Strategies, Grouping, and Ranging Patterns in Wild Chimpanzees. PhD thesis. University of Minnesota, St. Paul, MN.
  3. Sandeep Mane, CarsonMurray, ShashiShekhar, JaideepSrivastava and AnnePusey. (2005) Spatial Clustering Of Chimpanzee Locations for Neighborhood Identification, Proceedings ofIEEE International Conference on Data Mining, 2005, Houston, TX. (Extended version under review for IEEE Transactions on Knowledge and Data Engineering).
  4. Jin Soung Yoo and Shashi Shekhar, A Join-less Approach for Mining Spatial Co-location Patterns, the IEEE Transactions on Knowledge and Data Engineering (TKDE), Vol. 18, No. 10, October 2006.(Preliminary results appeared in proceedings of the IEEE ICDM 2005 and ACM-GIS 2004).
  5. CarsonMurray, Sandeep Mane and AnnePusey.(2006), Dominance rank influences female space use in wild chimpanzees (Pan troglodytes schweinfurthii): Towards an ideal despotic distribution, under revision for a journal in behavioral ecology.