A cost-sensitive technique for positive-example learning supporting content-based product recommendations in B-to-C e-commerce

► We address the single-class learning challenge that prevails in B-to-C e-commerce. ► We develop the COLPEL technique for online product recommendations. ► We evaluate the COLPEL technique with top reviewers' book ratings at Amazon.com. ► COLPEL outperforms all the benchmark techniques in terms of prediction accuracy and average F1 score. ► COLPEL benefits from cost-sensitive learning and the committee machine strategy.