Dr. Stephen Dipple supervised by Prof. Gyorgy Korniss and Prof. Bolesaw K. Szymanski defended the Ph.D. thesis on "Modeling Complex Human Behavior in Socio-Economic Networks" in Physics.

Stephen Dipple supervised by Prof. Gyorgy Korniss and Prof. Bolesaw K. Szymanski defended the Ph.D. thesis on "Modeling Complex Human Behavior in Socio-Economic Networks" in Physics. The thesis focuses on human behavior seen as a complex system of interacting individuals. These complex interactions can be de fined in terms of an encounter graph where people represent nodes and edges represent the micro-interactions in the population. Most behavior is not well modeled because of lack of complete information about the interactions between people. This thesis examines three complex social interaction environments. The fi rst model captures the dynamics of partner selection for mating. While this work can be applied to other types of scenarios, mating is the most popular scenario where two groups select partners from the opposing group. It is found that individuals can form assortative pairs based on their attractiveness even though a driving force for assortativity is not explicitly built into the model. In addition, more pairs can be matched from a population with a higher degree of selectivity at the cost of longer time required for all participants to fi nd partners. With a model for producing offspring based on the traits of their parents, populations go through multiple generations of pairing and reproduction. Surprisingly in this process, the distribution of attractiveness narrows and depending on the offspring variance forms a stable distribution away from the most desired attractiveness even for very selective populations. The second environment focuses on a Call Detail Record which catalogs the various communications between individuals using a particular cell phone carrier. A reciprocity measure can be constructed, which has the potential to show how embedded a user is within the cell phone network. Three measures were posed to account for various considerations such as noise, however there was a high degree of correlation between these measures. This measure is then compared to the list of users who failed to pay (aka. defaulted on) their cell phone bill. Indeed, reciprocity revealed some degree of correlation with users defaulting which signifi es that low reciprocity may indicate a low cost for leaving the cell phone network. The last project explores the fluctuations and interactions of the cryptocurrency markets and the associated social media outlets. Here, an in-depth model is presented which is capable of predicting not only daily changes in cryptocurrency markets, but also daily activity in various social media. We achieve this by examining correlations between the noise of these data sets. In this way, we are able to restrict the distribution of possible outcomes based on the assumption that correlations will persist during our prediction window. Our predictions farther into the future become increasingly inaccurate, however our method is reliable in predicting spikes and dips in the data even during long term predictions. This has greater signifi cance for cryptocurrency markets as the exact values of the market prediction is not as important as simply whether the market will increase or decrease. Our approach indeed has achieved an impressive performance compared to the random prediction and our baseline measure for predicting whether a market will go up or down the following day.