Complex system simulation: agent-based modeling and system dynamics

COCCO, LUISANNA

2013-04-23

---

Abstract

This thesis deals with some simulation based approaches used to study software market and software development. Nowadays, the term Software as a Service is everywhere and is described as the future of software. SaaS, also called On-Demand Software, is a software application delivery model that together with Commercial Open Source Software another pricing approach is slowly gaining ground. Indeed, in recent years, traditional software also called On-Premise software appears overpriced, user's willingness to buy it is decreased and therefore the purchase preferences are moving from traditional pricing models to new pricing approaches. To study these new pricing tendencies, different models have been realized by using two of the most common numerical techniques: Agent based Modeling and System Dynamics. With agent based modeling two business models have been realized: a model to study the competition among CRM On-Premise and On Demand vendors and another model to study the competition among CRM On-Demand vendors offering CRM products, with and without source code availability. Our goals are to propose business models to analyze and study the CRM software market, and to propose a useful tool to forecast business winning strategy and investment and pricing business policies. Instead, with system dynamics a tool for highlighting how a Global Software Development environment on the Cloud Platform may facilitate GSD with respect to an environment set up On Premise has been realized. All these models are based on many insights from literature and market analysis. However, concerning the business models, this is the first time that the software market has been modeled using heterogeneous agent model and detailing investment and pricing policies of firms and purchase preferences of customers, and consequently building the model on existing scientific knowledge has not been simple. In addition, lack of experimental data to initialize or validate the models clearly limits the validity of the models, and for this reason the future main objective will be to validate the model using real enterprise data.