Data-Driven Simulation-Based Analytics

This chapter will introduce a novel data-driven simulation-based approach to study a dynamic system. Classification and regression trees a data mining method – is utilized to extract non-linear dynamics of the system based on “big data” collected over a long period of time. Kernel density estimates are used at the terminal nodes of the tree to model the probability distribution of the response variable. Then the simulation model repeatedly samples responses from the kernel density estimates based on the present state of the system to mimic the evolution of the system. This way of simulation is completely data-driven and parameter free, it avoids misrepresentation of the system. Application of this method is illustrated with two diverse applications. In the first application, nurse-activity is simulated based on a big data set collected at a northeast Texas hospital. Five tree structures are developed: (a) four classification trees from which transition probabilities for nurse movements are determined, and (b) a regression tree from which the amount of time a nurse spends in a location is predicted based on factors such as the primary diagnosis of a patient and the type of nurse. In the second example, the dynamics of the financial market movement in the United States is estimated by simulating the Standard and Poor’s 500 Index. Unlike any other indices, the Standard and Poor’s 500 Index (S&P500) is the most commonly used index in the New York Stock Exchange to understand the market in the United States. It was introduced in its current form in March, 1957. The 500 publicly traded companies that make up the index are chosen by a committee to best reflect the overall market in the United States. The data set included the daily movement of financial markets in seven countries in Asia and Europe in relation to the daily movement of the S&P500. Trees also utilized data on the currency exchange rates to capture the financial dynamics between the US and other countries. The simulation model repeatedly samples from four trees to know how the opening and closing values of the S&P500 move in tandem with the other markets.