Machine Learning Model for Address Verification

This project focused on developing a machine learning model to identify wrong addresses before the first delivery attempt for Loggi. Initially, the delivery failure rate due to incorrect addresses was 5%, which was later found to be improved with proactive intervention.

By analyzing successful deliveries that required address corrections, the model utilized city, neighborhood, latitude/longitude, and client data to achieve an 80% accuracy in identifying incorrect addresses. This model allowed human intervention to verify and correct predictions, reducing the failure rate from 5% to about 3%. This proactive approach significantly enhanced the efficiency and reliability of the delivery process, saving both time and operational costs.