Temporal Resolution vs. Accuracy of Rhesus Macaques’ Social Network Based on RFID Tracking Data Open Access

Wang, Yizhou (Spring 2023)

Permanent URL: https://etd.library.emory.edu/concern/etds/tb09j716v?locale=en


The adaptation of the tracking device enables primate studies to obtain more comprehensive and longitudinal behavior data of the animals. Nevertheless, such data often comes with enormous computational challenges. In the previous relevant studies, it was discovered that data obtained at a very short timescale is comparable with the actual human observational data (Gelardi et al., 2020). In addition to the computational challenges, few studies focused on using longitudinal tracking data to generate and infer the social network of the rhesus macaque. Therefore, this study is aimed to evaluate the accuracy vs. temporal resolution for generating the social network of Rhesus Macaque with 154 days of data collected from 2021 March 4th to 2021 August 4th. Through investigating the spectral clustering result and the Euclidean distance of the proximity matrices generated from different averaging-over durations and the correlation of them with the actual matrilineal genetic similarities, it was found that averaging-over durations of 14 or 28 days are potentially optimal for generating a stable social network that can represent the entire period between March 4th, 2021 and August 4th, 2021. 

Table of Contents

Introduction 1

Background 2

The Significances of Rhesus Macaque in Psychological and Biomedical Studies 2

Social Structure of Rhesus Macaque 4

Proximity Measurements and the Tracking Device Data 5

Weighted Adjacency Matrix and Similarity Matrix 6

Laplacian Matrix 7

Spectral Clustering 8

K-means Clustering and K-medoids Clustering 9

Rand Index and Modified Rand Index 10

Euclidean Norm 12

Methods 12

Data Collection 12

Preliminary Processing and Correction of the Data 13

Proximity Calculation 13

Spectral Clustering 15

Selecting The Optimal Number of Clusters k 15

Spectral Clustering With k = 6 19

Selecting the Optimal Window-away Distance 19

Selecting the Optimal Averaging Duration 22

Selecting the Optimal Window Within the Optimal Window-away Distance for Visualization Purposes 24

Euclidean Distance of Matrices 26

Selecting the Optimal Window-away Distance 26

Selecting the Optimal Averaging Duration 27

Comparing the Results With the Actual Matrilineal Genetic Tree 28

Results 30

Background 30

Results of Spectral Clustering Evaluated by Modified Rand Index 31

Results of Euclidean Norm of the First Eigenvectors 37

Social Network Generated Based on the Selected Data 38

Discussion 41

Limitations and Future Directions 48 

About this Honors Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
  • English
Research Field
Committee Chair / Thesis Advisor
Committee Members
Last modified

Primary PDF

Supplemental Files