Abstract:
Factorization Machines (FMs) help recommender systems overcome cold-start and data sparsity by incorporating side information. Traditional FMs use the inner product to model second-order interactions between feature vectors. Inner products violate feature vector triangle inequality. It performs poorly because it cannot capture fine-grained attribute interactions. FMs have improved performance by replacing the inner product with the euclidean distance. However, previous FM methods, including those with the euclidean distance, focused on attribute-level interaction modeling, ignoring the critical intrinsic feature correlations inside attributes. They fail to model real-world data’s complex and rich interactions. This paper proposes an FM framework with generalized metric learning to better capture feature correlations. Based on this framework, we present Mahalanobis distance and deep neural network (DNN) methods to model linear and non-linear correlations between features. We also simplify model functions efficiently. Our framework significantly outperforms several state-of-the-art baselines on several benchmark datasets. Our method overcomes cold-start and data sparsity issues in recommender systems.
Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.
Did you like this final year project?
To download this project Code with thesis report and project training... Click Here
