Performance analysis of unsupervised learning algorithms in the group classification problem for customs risk management

Abstract

Currently, the revision of legislation on international express shipments (small packages), commonly known as import via electronic commerce, is in vogue. In view of this, there is a need to propose a policy, accompanied by a review of legislation and the proposal of more efficient control actions, to regulate, control and, eventually, tax the import of small parcels. In this sense, artificial intelligence techniques can help identify characteristics that may be important for more efficient control actions. Joint meetings will be held with representatives knowledgeable about operational procedures related to problems of monitoring international express shipments. Clustering or cluster analysis is an unsupervised learning problem. It involves automatically discovering natural groupings in data. Unlike supervised learning (like predictive modeling), clustering algorithms just interpret input data and find natural groups or clusters in the feature space. Evaluation metrics will be defined and used to decide the best combination of the model set and solution techniques to be employed. To this end, case studies generated from operational data will serve to better understand the solutions proposed in each model/method combination and which are the most appropriate.