A Text Classification Model to Identify Performance Bonds Requirement in Public Bidding Notices

Abstract

A performance bond is a means of guaranteeing that a product that will be delivered by the seller in a timely and workmanlike manner. To verify whether a performance bond has been laid down in a bidding notice, an assessment through an internal auditing process is made periodically, but it can be costly and take longer than the available time to conclude an audit engagement. We propose to explore and apply algorithms to create a model able to identify bidding notices that demand performance bonds, to make the assessment process more efficient. We applied Four different classification algorithms (SVM, kNN, Random Forest, and Naive Bayes) on two different vector space representations (term frequency and term frequency inverse-document frequency). Random Forest with term-frequency produced the best model, which achieved-score of 0.933 on the test set. The promising results show the model created could help decrease the time and effort spent on verifying the performance bonds requirement on bidding notices.