From the course: Complete Guide to Generative AI for Data Analysis and Data Science

Introduction to simulations

From the course: Complete Guide to Generative AI for Data Analysis and Data Science

Start my 1-month free trial Buy for my team

Introduction to simulations

“

- [Instructor] Simulations are an important part of data analysis and data science. Now, basically a simulation is a way of modeling a sequence of events. Now, typically there are multiple events and multiple possible states, and so it's this combination of multiple events and multiple possible states over repeated studies or repeated simulations that really lend insight into complex phenomenon. Now, some common characteristics of simulations are we often use them when we're dealing with uncertainty. So we may not have data to help us understand a particular phenomenon. So, for example, if we want to know, you know, what's the expected number of products we're going to sell in a particular region in a particular quarter, well, we can maybe look at the previous quarter or that quarter from last year and probably get a good idea. Now, complex interactions is another indicator or property that we commonly see in problem spaces where simulations work well. And you can think about things like infectious disease spreads. Also, there are cases where we just don't have analytic solutions. You know, there might be equations for solving particular problems. Or if you're good at calculus and you like working with differential equations and you can model something as a differential equation, you can probably come up with some insights about the behavior. But we don't always have easy access to an analytic solution. So sometimes we have to run simulations to see what the outcome is. And then, finally, we just don't have real-world data available, or the data that we have is insufficient. Again, this can be because there might be changes in things like we might have a lot of historical data about, say, power demand for, a demand for electricity in the summer months in different parts of the country. But maybe there are significant changes. People have started buying electric vehicles in this particular region, or the pace at which people have purchased electric vehicles has really accelerated, so demand goes up. But maybe there's some other changes. There's also climate change, and maybe temperatures are going up, so there may be more demand for air conditioning and other kinds of environmental controls. So looking at historical data might actually kind of lead us to underestimate certain kind of demands. So we may have historical data, but we feel like it doesn't fit. Now, we can use that historical data to start estimating parameters and distributions of certain variables and get a sense of bounds around uncertainty. But if we want to understand how multiple uncertain factors are influencing each other and what the total impact of those multiple factors are, simulations often are a good way to go. These are a tool that we turn to.

Contents