Privacy-Friendly Delivery Plan Recommender

When planning the distribution of parcels, postal services often face different problems. One of them is delivering a parcel to a customer's home while making sure the customer is there to receive it. Undeniably, a successful first-try delivery minimizes monetary costs and is less resource-consuming. The goal of this project is to recommend delivery plans to postal services, in order to achieve a better success rate for home deliveries on the first try. A dataset of three years of deliveries of a postal company located in Switzerland is analyzed, to create appropriate features for classifying customers based on their past deliveries. The K-Means algorithm is applied to achieve this classification. By only using anonymized information about the customers' past deliveries, which is information already owned by postal services, it avoids invading the data privacy of these customers while still providing a viable solution.