Abstract Data scarcity is a problem for recommender systems, and cross-domain recommendation has been shown to be an effective solution. Previously, this strategy has worked well. We've done some work in the past to improve Deep Learning IA-CNN-based Personalized Product Recommendation Systems using Hybrid Deep Learning-based Hybrid Techniques. An approach to cross-domain deep neural networks is presented in this paper, which utilizes both deep neural networks and cross-domain recommendations (CD-DNN). With the help of user reviews and product metadata, CD-DNN can accurately predict ratings for many different kinds of products. If you want CD-DNN to be able to accurately predict future ratings of users or items in your target domain and other source domains, you'll need to train it on data from both of those places. By maximizing the accuracy of prediction predictions, several parallel neural networks are trained to learn the latent factors for both user features and item features, as well. When CD-DNN and other domain features are combined, a single latent space mapping for user attributes is produced. By doing so, the network's users benefit from improved performance. The proposed CD-DNN outperforms other current recommendation approaches, and it also addresses data scarcity by incorporating data from a variety of sources, including the Amazon datasets.
Alan : Mühendislik
Dergi Türü : Uluslararası
Benzer Makaleler | Yazar | # |
---|
Makale | Yazar | # |
---|