High Performance supervised learning for recommendation and predictionTechnology #2017-075
Questions about this technology? Ask a Technology Manager
- Prof Chua Tat Seng NExT Centre for Extreme Search E-mail: firstname.lastname@example.orgExternal Link (www.nextcenter.org)
- Managed By
- Ms Asha Srinivasan (email@example.com) Associate Director (65)65161057
- Patent Protection
- US Patent Pending
AFM: High performance supervised learning method for recommendation and prediction
Factorization machines (FM) are a new model class that combine the advantages of support vector machines, one of the most popular predictors in machine learning, data mining and factorization models. FMs are mainly used for supervised learning under sparse settings and good for general prediction. Despite its effectiveness, FMs can be hindered by its modelling of all feature interactions with the same weight, as not all feature interactions are equally useful and predictive.
Attentional Factorization Machine (AFM) is a novel model which improves FM by discriminating the importance of different feature interactions and learns the importance of each feature interaction from data via a neural attention network. More importantly, the importance of a feature interaction is automatically learned from data without any human domain knowledge.
Stage of Development
Analytical and laboratory studies to validate analytical predictions
AFM will be particularly effective for prediction tasks involving categorical predictor variables, such as user/item IDs, attributes, tags and categories. As an example, it can be used in the ranking engine of recommendation systems and click-through-rate (CTR) prediction of online advertising systems.
The importance of a feature interaction is
automatically learned from data without any human domain knowledge.
AFM consistently outperforms the
state-of-the-art deep learning methods with a much simpler structure and fewer
model parameters, leading to better performance.
AFM provides insights into which feature
interactions contribute more to the prediction. This greatly enhances the
interpretability and transparency of FM, allowing practitioners to perform
deeper analysis of its behaviour.
Patent pending. Available for licensing.
Prof Chua Tat Seng
NExT Centre for Extreme Search
ILO Ref No: 2017-075