We propose a method to improve recommender systems  via a Dual (user and item) Training Error based Correction approach (DTEC).  The proposed method is applied to the Synthetic Coordinate  Recommendation system (SCoR) and to other state-of-the-art recommender systems. see Fig. 1.  

• In this work, we introduce a second stage (DTEC) after a Recommender System execution (e.g. SCoR), that corrects the recommendations of the Recommender System taking into account the error on the training set between users and items and optionally their proximity in the synthetic Euclidean space of SCoR.

•  These corrections can be performed in user and item view point, so finally a dual system is proposed to efficiently combine both corrections. 

• DTEC computes a model that makes zero the recommendation error on the training set, and then applies it on the test set to improve SCoR’s predictions. 

• The proposed DTEC approach is applicable on any model-based recommender system with positive training error like SCoR, potentially increasing the accuracy of the recommendations. 

Experiments - Downloads

• • You can download the matlab code of the DTEC method proposed in [1,4]. 

  • The proposed method is available on codeocean : https://codeocean.com/capsule/8714785 

  • You can download the datasets of the method proposed in [1,4]  from (.zip).

Related Publications

[1] C. Panagiotakis, H. Papadakis, A. Papagrigoriou and P. Fragopoulou, Improving Recommender Systems via a Dual Training Error based Correction Approach,   Expert Systems with Applications, 2021. 

[2] C. Panagiotakis, H. Papadakis, and P. Fragopoulou, A user training error based correction approach combined with the synthetic coordinate recommender system,   UMAP, 2020. 

[3] H. Papadakis, C. Panagiotakis and P. Fragopoulou, SCoR: A Synthetic Coordinate based Recommender System, Expert Systems with Applications, vol. 79, pp.8-19, 2017. 

[4] 39. C. Panagiotakis, H. Papadakis, A. Papagrigoriou, and P. Fragopoulou, DTEC: Dual Training Error based Correction approach for Recommender Systems, Software Impacts, 2021.