Page 286 - Data Science Algorithms in a Week

Agent-Based Modeling Simulation and Its Application to Ecommerce 267

There is one transition from InRepayment state and this has two branches. One of the
branches leads to FullyPaid while the other to the InDefault state and then to Exit where
the borrower leaves the system on charge off. The decision at the TestDefault branch is
made internally and stochastically. The average amount of capital and interests that is
repaid, recovered or lost when a borrower defaults is also reflected.
LC, which acts as a central dispatcher, broadcasts requests for borrower loans to all
lenders. For simplicity, LC is modeled as a function call that responds to requests. LC
listens for messages from the borrower as well as the lender side and manages the
transaction completion on behalf of the agents. LC inserts a message in the queue and
notification is broadcasted to borrowers and lenders. BorrowerAB and LenderAB
represent borrower and lender agent classes. The communication instances used in the
model are summarized below:

1) Screening request: a message arrives from the borrower and lender requesting
screening.
2) Interest rate generation: the LC generates an interest rate and communicates it to
the borrower.
3) Borrower decision on interest rate: based on the risk profile, the borrower decides
to accept or reject the generated interest rate.
4) Lender’s decision on interest rate: the lender decides to fund a particular
borrower with an associated interest rate based on its risk profile.
5) Payment: payments are communicated to LC and in turn to the lender.
6) Default: the borrower leaves the platform and the lender and borrower returns are
updated.
7) Fully paid: a message from the borrower and lender deciding if to go back as
potential customers or they can choose to leave the system.

It is assumed that participants are sensitive to ads and word of mouth (WOM). The
WOM effect is the way new users are persuaded to purchase a product or adopt a service.
Consumers persuade others to adopt a service or buy a good often using word of mouth.
Each participant’s adoption time differs. In this system, customer satisfaction measured
by response to WOM and results from satisfactorily completed loans. Hence, it is
expected that as more customers default, the WOM decreases. A consumer contacts an
average of number people in a month i.e., a specified contact rate. Agents in the system in
turn contact each other and influence potential borrowers to sign up for the service.
Space and request queue management are defined within the Main with space and
layout requirements configured in the Environment object contained in the Main. A value
of 1000 each was assigned as initial number of borrowers and lenders in the system.