Knock knock
Nidhi R. Arora · 2016
Address Cleansing is a very challenging problem, in particular, for geographies where there is a lot of variability in address nomenclature. This paper explains how a machine learning classifier is used to build a model that not only identifies a valid address, but also assigns a confidence score to it. This confidence score is then used to perform address cleansing and correction. The experiments have been conducted using Mallet API in Java and experimental analysis shows 75% accuracy.