TMRM: Two-Stage Multi-Task Recommendation Model Boosted Feature Selection

For recommendation systems, full learning from features is critical to improving system accuracy. Most recommendation systems collect a large number of features to improve the recommendation accuracy, but they ignore the importance of extracting important feature combinations. Even if some systems manually measure important features based on experience, they cannot automatically select important feature combinations, that is why they can not be used to guide model training. In this paper, a Two-stage Multi-task Recommendation Model(TMRM) is proposed which aims to automatically select important feature combinations from massive features, and it also contributes to achieving better recommendations through a combination of tree-based model and neural network. Extensive experiments on two large public data sets are conducted on TMRM, and the results demonstrate the superiority of our proposed method over state-of-the-art solutions on performance of recommendation systems.

[1]  Alexandros Karatzoglou,et al.  Recurrent Neural Networks with Top-k Gains for Session-based Recommendations , 2017, CIKM.

[2]  Guokun Lai,et al.  Explicit factor models for explainable recommendation based on phrase-level sentiment analysis , 2014, SIGIR.

[3]  Yue Yin,et al.  Explainable Recommendation via Multi-Task Learning in Opinionated Text Data , 2018, SIGIR.

[4]  Thomas Brox,et al.  Pixel-Level Encoding and Depth Layering for Instance-Level Semantic Labeling , 2016, GCPR.

[5]  Yunming Ye,et al.  DeepFM: A Factorization-Machine based Neural Network for CTR Prediction , 2017, IJCAI.

[6]  Roberto Cipolla,et al.  Multi-task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semantics , 2017, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[7]  Tat-Seng Chua,et al.  TEM: Tree-enhanced Embedding Model for Explainable Recommendation , 2018, WWW.

[8]  Heng-Tze Cheng,et al.  Wide & Deep Learning for Recommender Systems , 2016, DLRS@RecSys.

[9]  Olfa Nasraoui,et al.  Using Explainability for Constrained Matrix Factorization , 2017, RecSys.

[10]  Tat-Seng Chua,et al.  Neural Factorization Machines for Sparse Predictive Analytics , 2017, SIGIR.

[11]  Aapo Hyvärinen,et al.  Noise-contrastive estimation: A new estimation principle for unnormalized statistical models , 2010, AISTATS.

[12]  Joaquin Quiñonero Candela,et al.  Practical Lessons from Predicting Clicks on Ads at Facebook , 2014, ADKDD'14.

[13]  Xu Chen,et al.  Learning to Rank Features for Recommendation over Multiple Categories , 2016, SIGIR.

[14]  Qian Zhao,et al.  GB-CENT: Gradient Boosted Categorical Embedding and Numerical Trees , 2017, WWW.

[15]  Olivier Chapelle,et al.  Field-aware Factorization Machines in a Real-world Online Advertising System , 2017, WWW.

[16]  Carlos Guestrin,et al.  "Why Should I Trust You?": Explaining the Predictions of Any Classifier , 2016, ArXiv.

[17]  Paul Covington,et al.  Deep Neural Networks for YouTube Recommendations , 2016, RecSys.

[18]  Diyi Yang,et al.  Combining Factorization Model and Additive Forest for Collaborative Followee Recommendation , 2012 .

[19]  Carlos Guestrin,et al.  Anchors: High-Precision Model-Agnostic Explanations , 2018, AAAI.

[20]  Fan Zhu,et al.  RSLIME: An Efficient Feature Importance Analysis Approach for Industrial Recommendation Systems , 2019, 2019 International Joint Conference on Neural Networks (IJCNN).