A Comparative Study of Machine Learning Techniques for Automatic Product Categorisation

Abstract

The revolution of the digital age has resulted in e-commerce where consumers’ shopping is facilitated and flexible such as able to enquire about product availability and get instant response as well as able to search flexibly for products by using specific keywords, hence having an easy and precise search capability along with proper product categorisation through keywords that allow better overall shopping experience. This paper compared the performances of different machine learning techniques on product categorisation in our proposed framework. We measured the performance of each algorithm by an Area Under Receiver Operating Characteristic Curve (AUROC). Furthermore, we also applied Analysis of Variance (ANOVA) to our results to find out whether the differences were significant or not. Naïve Bayes was found to be the most effective algorithm in this investigation.