An Application of Recommender System with Mingle-TopN Algorithm on B2B Platform

Personalized Recommender systems have been widely used on many large websites in various business fields to tackle with problems like excessive information and information overload. For the purpose of improving user experience and stickiness, this study proposes a personalized Recommender system for facilitating the users in the B2B ECommerce platform to retrieve and extract information. The proposed Mingle-TopN hybrid Recommendation algorithm combines collaborate filtering and content-based filtering, to improve the effectiveness and efficiency of the recommendation. The Mean Absolute Error (MAE) is used as the metric to evaluate the developed Recommender system. Results show that: when the number of items exceeds 600, the proposed method trends to achieve a stable performance with a MAE of 0.73. These experiments validate that the improved recommendation algorithm is better than the Pearson-TopN and Cos-TopN by 10.61 % and 13.45%, respectively.