1. What is simulation?
In simulation, we imitate real-world phenomena in computer form (similar to computer games). This simulation allows us to evaluate different strategies on a computer. In many cases, doing an experiment using computer simulation is cheaper, less disruptive and safer than doing it directly in the real world. However, in order to obtain valid simulation results, we need to follow a series of scientific steps (known as simulation modelling methodology). The methodology is partly influenced by how modellers view the real world (known as a simulation modelling paradigm).
2. What is business analytics?
It is a process of transforming data into actions through analysis and insights in the context of organizational decision making and problem solving (Liberatore & Luo, 2011). There are three types of business analytics: descriptive analytics (focusing on the understanding of what has happened), predictive analytics (focusing on estimating what will happen) and prescriptive analytics (focusing on estimating what should happen). Simulation can be used in all three types. However, business analytics professionals typically use simulation for predictive analytics and combine it with an optimization model for prescriptive analytics.
3. What are the common dynamics simulation modelling paradigms in business analytics?
Discrete-event simulation: it views the world as a set of interacting processes. Each process may consist one or more activities. Entities join the system and spend time in one or more activities before leave the system. An activity may require one or more resources before it can start. If one or more resources are not available, an entity may need to queue before it spends time in the activity.
Agent-based simulation: it views the world as a set of interacting agents (or actors). Hence, it is suitable for analysing the behaviour of agents at a micro level, the resulting behaviour at an aggregate level, and interactions between the two levels over time. ABS has become increasingly popular due to increasing computer power and, most importantly, the availability of more detailed data, such as: online interactions (from Web logs), consumer behaviour (from various membership cards), spending behaviour (from credit/debit cards transactions), and travel behaviour (from Oyster cards or other similar travel cards).
System dynamics: it views the world as a set of stocks and flows. It models a system at the aggregate (population) level. Hence, it allows wider model boundary which is good in analyzing the interaction of multiple feedback loops in a system.
4. How do we use simulation to solve real-world problems?
We solve a real-world problem by developing a simulation model that represents the most important elements of the real-world system. A properly designed experiment is conducted using the simulation model to get some outputs that need to be interpreted to produce actionable insights.
In solving real-world problem, simulation is often combined with other tool, for example:
Optimization-via-Simulation (or Simulation Optimization) which combines simulation and optimization models where the objective function in the optimization model is estimated using a simulation model.
Simheuristics which combines a simulation model and heuristics.
Multi-fidelity model which combines simulation optimization with a faster low-fidelity model to guide the high fidelity simulation optimization model.
Professor Stephan Onggo 