Project Overview

                This project focuses on Challenge #2: Expedite the Scoring Process & Expand Onboard Families.
                For the first task, Classification and Regression Tree (CART) method is introduced. The decision tree is used to predict how likely an applicant will adopt a child in a permanent manner by examining his/her profile. Taking the probability as the “score”, this method can significantly accelerate the relevant process in an agency.
                For the second task, an unfolding multidimensional scaling (MDS) model is adopted. This method helps us visualize what type of children is more attractive to a certain type of families by mapping these two-mode data onto a same plot. The agency can then match families with the closest kids showing on the plot in order to raise their interest in and patience with the adoption process and reduce the attrition rate in the duration.
                At last, two interactive tools based on these two methods are designed for Foster Care agencies to quickly score an applicant and match an onboard family with a kid.

Methods

Selecting Datasets:
                 Adoption and Foster Care Analysis and Reporting System (AFCARS), Adoption File FFY 2015 and Foster Care FFY 2015 by the Children's Bureau (St = “NY”)
Generating Variables:
1. For Decision Tree: 23 variables in total