Development of machine learning models for classification of tenders based on UNSPSC standard procurement taxonomy

Nigerian public procuring entities are gradually transitioning from manual-based procurement processes to digital-based processes. One of the key processes being automated is the notification of tenders which traditionally has been done through newspapers. Given the growing volume of digital advertisements, it is very imperative to automate the process of classification of tender titles/descriptions into appropriate categories using standard procurement taxonomy such as the United Nations Standard Product and Service Code (UNSPSC). Natural language processing (NLP) methodology was applied to automatically classify tender titles/descriptions into appropriate UNSPSC. Multiple machine learning algorithms were employed to develop the classification models. Given a tender title/description, the models can predict the correct code to apply at the segment, family, class, and commodity levels of the UNSPSC. The algorithms with the best performance under the train-test split validation and K-fold cross-validation methods are support vector machines.