This experiment is set to run with a high frequency of offline applications. Moving to the Offline applications prevalence MC1 experiment, we can find another of the new features, 2D histograms.
Try some different ranges for the application ratio and see how the credit approval rates are affected. These iterations run for the number of replications specified, each replication having a different seed value that affects the parameters inside the model, for example, the customer arrival rate. A random distribution of values is created, one application ratio figure for each iteration. The ratio of applications is bound by the two values entered in the App ratio range. This experiment allows us to randomly set the ratio of applications coming from people in a bank branch office or online. To try Monte Carlo 2nd Order, select the Offline ratio variation sensitivity experiment in the sidebar. Not just randomizing internal parameters of a model, as with a Monte Carlo 1st Order experiment. 2nd Order Monte Carlo allows us to randomly set both internal and input parameters according to a probability distribution. We can use Monte Carlo 2nd Order experiments to solve these problems. Some experiments can have input ranges that do not have clear steps. For example, this scenario shows the number of abandoned applications is lowest when there are four bank clerks and five analysts (see the blue 3D surface graph in the lower left).Ĭlick to animate: New box plots and 3D surface graphs.
The 3D surface graphs allow us to see the interaction of two variables. The line within the box marks the median, 50% of the applications took less than this time and fall between it and the lower line outside of the box (the lower whisker). The data is split into quartiles, with 25% of the applications in each. The box plots indicate the range of time applications were in the system. Scrolling down will show you the progress of the experiment and then the output, where you will find more of the new features.īoth the new box plot and 3D surface graphs appear at the bottom of the experiment outputs. Set the parameters and the number of replications you would like and then hit Run in the top menu bar. By running each combination a number of times, you can test the robustness of the model to get greater confidence in the setup.Ĭlick to enlarge: Staff capacity variation experiment inputs. A seed is a number which is used to determine randomly generated values such as the arrival rate of customers or the processing time of forms for an employee. Each combination can also be replicated, meaning the experiment runs several times with different seeds. An experiment is run for every combination of both ranges, this provides the variation. We can vary the inputs and change the number of bank clerks and analysts by setting the range and step change. This time, we will use replication because we are unsure of the effect of variables such as the customer arrival rate. Choose Variation, and not Variation with Replication, if this is the case when you are modeling. It is worth noting that replication is not needed when you are certain your model is not significantly affected by randomness. This experiment provides an example of Variation with Replication. Open the model and select Staff capacity variation from the sidebar (on a mobile device, you may need to click the sandwich icon in the top left to expand the sidebar). In this model we’ll look at the new experiments and graphs.
By using the results of multiple model runs, we can determine approval rates and find an optimal number of bank employees.
The verification process involves three stages: scoring, personal review, and credit rating inspection. The Consumer Credit model describes the life-cycle of consumer credit applications, from both a branch office and online. In Public models find the Consumer Credit model, use search if you don’t see it immediately.Ĭlick to enlarge: The Consumer Credit model and sidebar. Let’s take a closer look at these new features. Outputs now include Box Plot, 2D histogram, and 3D surface graphs. There are also enhancements to help you communicate your models more widely and clearly. Among them, Monte Carlo 2nd Order, Variation, and Variation with Replication experiments. A host of new features were added to the AnyLogic Cloud recently.