Page 287 - Data Science Algorithms in a Week

268 Oloruntomi Joledo, Edgar Gutierrez and Hatim Bukhari

An advantage of object oriented ABM is that we can look deeper into each object –
borrower or lender – and view its state and variable values. The following are some
inputs used in calibrating the agent based model (Figure 5):

 The number of borrowers is initialized to 1000.
 A random borrower can request anywhere from $1000 to $35,000 and based on
his profile.
 The contact rate is kept at 1.5% to prevent the number of new agents entering the
system from growing too large.
Simulation experiments help to facilitate systematic and quantitative analysis on the
effects of factors of interest. Simplifying modeling assumptions adopted for this study
include:

 A given lender is attached to a given borrower.
 Agents leave after they complete payment.
 A borrower has an option to return to the state of potential borrower.
 Agents who default must leave the system.
 Probability distributions are used to generate the agent profiles.
 Arrival patterns of borrowers and lenders are based on LC user arrival rate.
 Term of loans is either 36 months or 60 months and the choice follows a
probability similar to real data.
 State transitions are instantaneous and time durations are factored into the
timeout triggered property.

Figure 5. Agent-based simulation interface.