This project integrates the logic of a KAPSARC Toolkit for Behavioural Analysis (KTAB)-style model of collective decision-making processes (CDMPs) with machine learning to develop a hybrid approach that can be used to anticipate politically-driven outcomes for population-scale analytic questions. If successful, this will extend the applicability of KTAB to include the decision-making of families, households, and actors at the scale of a market or country’s population. The first proof of concept study applies a custom deep-learning neural network to predict household vehicle purchasing decisions in an entire market. The proof of concept has two primary research goals to test the viability of this project:
1) To demonstrate that the CDMP underlying KTAB can process objective data, not just expert opinion.
2) To demonstrate that embedding a CDMP in a neural network improves performance.