Trust Based Recommendation Framework Using Cross Domain Information

Abstract

Recommender system has become a vital and indispensable component in the modern Internet industry. Since the 1990s, recommender systems have gained popularity in education, business, tourism, healthcare, entertainment, and culture. Many e-commerce sites use recommender systems as business tools to increase their sales productivity and help their customers find suitable products. Due to sparse rating and lack of historical information, the traditional recommender systems cannot generate effective recommendations. Over the years, cross-domain and trust-based recommendation systems have proven to help solve issues about data sparsity and cold start. Although,cross-domain recommendations have been improved pervasively; however, the challenge of employing several influences, including time, trust, and location, remains unresolved. Many approaches have been researched on how the recommendations are generated for users; however, only a few work, considering the time factor that is at what instant of time a user may show interest in a particular item. Inspired from this fact, we define a new problem that targets recommendation by considering three main factors time, trust, and location. In this research, we solved a user cold-start problem in trust-based recommender systems using cross-domain information. Our key contributions to personalized recommendations are mainly divided into three folds: First, we proposed a Trust-Aware Cross-Domain Temporal Recommendations (TrustCTR) model that tackles the problem of temporal dynamics and user cold start. TrustCTR predicts the items’ ratingsabout an active user from the most recent time by modeling preference drift on temporal basis by following the cross-domain scenario of ‘No Overlap’ in single-target cross domain recommendation. We performed experiments on public datasets Ciao and Epinions. We used these datasets in cross-domain forms, such as the categories of Ciao as the source domain and Epinions as the target domain. We selected five different domains, having a higher proportion of rating sparsity, for observing the performance of our approach using MAE, RMSE, and F-measure. Our approach is a viable solution to the cold start problem and offers effective recommendations. We also compared TrustCTR with state-of-the-art methods; the model generates satisfactory results. Secondly, we investigated the feature representation learning via autoencoder and proposed a model called Trust-Aware Cross-Domain Deep Neural Matrix Factorization (TCrossDNMF) based on a deep neural architecture that models the complex and nonlinear relationships between users and items. The working of TCrossDNMF model is divided into four main steps: i) Features learning that learns the users’ features using a latent factor model and then finds the similarity between users of source and target domains. As the users are shared between two domains, the proposed model learns the common information and transfers the knowledge from a source to a target domain. ii)Ranking that finds set of similar users (neighbors), and then filters out the dissimilar users based on similarity threshold 𝜃, and then generates a bipartite trust graph from these reduced set of users and find trustworthy neighbors for an active user. iii) Weighting computes the trust degree between an active user and his or her top-k neighbors. iv) Prediction that trains the TCrossDNMF model using multilayer perceptron (MLP) and generalized matrix factorization (GMF) by representing user item interactions in higher dimensions and ensembles the GMF and MLP with trust information for rating prediction. We evaluated a model on a real dataset collected from a popular e-commerce retail service ‘AliExpress’. We used categories available in ‘Ali Express’ as a source domain and a target domain. For observing the performance of proposed model, we took three domains which have higher ratio of sparsity. The proposed model is evaluated using MAE, RMSE, and F-measure metrics and compared with baselines. The experiments show that TCrossDNMF is a viable solution for the mentioned problem with significant improvements in results, in dual-target cross domain recommendation. Thirdly, we investigate the challenge of biasness in textual feedback and user cold-start problem and present a model that considers both item and user unbiased preference information under a unified deep neural framework that can introduce personalization and collaborative effects based-on cold-start recommendations. We demonstrate a Trust-Aware Spatial-Temporal Activity based Denoising Autoencoder (TSTDAE) model that generates top-N recommendations about an active user with best items’ recommendation time in the cross-domain scenario of ‘User Overlap’ in single-target cross-domain recommendation. The TSTDAE model work is five-fold: i) Filter out the users’ biased profiles based on sentiment analysis. ii) Learn the features using autoencoder and then perform clustering among the users of source and target domains and discover the best neighbors. iii) Compute the trust and ratio of preference bias between active users and their neighbors and grade the neighbors based on unbiased preferences iv) Project the best time for recommending the items to an active user v) Generate the top-N recommendations for an active user. The TSTDAE model is evaluated on ‘AliExpress’ data and utilizing the available categories in cross-domain form. The performance of TSTDAE is observed over four domains and is evaluated using Precision, Mean Absolute Precision (MAP), Normalized Discounted Cumulative Gain (NDCG), and Hit Ratio (HR) and compared to seven baseline algorithms. Extensive experimental results on the ‘AliExpress’ dataset verify the effectiveness of TSTDAE over other competitors.