Colaborative FIltering recommendation engine using nearest neighbours association. Designed as an internal component of BooXchange e-commerce platform.
Visualization engine in action.
Algorithm used is modified, weighted voting K-NearestNeighbors. Online learning and multilabel classification is used. Nonparametric algorithm was chosen due to lack of prior customer data.
Visualization tool enabled customer behavior analysis by displaying customer segments (generated by clustering algorithm) and associated products bought by each customer. Dimensionality reduction was used to plot customer colaborative filtering graph in 2D.