| Autor: |
Sanjeev, Dhawan, Singh, Kulvinder, Craciun, Eduard-Marius, Rabaea, Adrian, Batra, Amit |
| Předmět: |
|
| Zdroj: |
Neural Processing Letters; Dec2023, Vol. 55 Issue 7, p9409-9434, 26p |
| Abstrakt: |
In modern times, next-cart recommendation (NCR) has appeared as an excitable subject of research among researchers and is dominant in e-commerce. In NCR, a customer buys a set of products (termed a cart) in a particular interval of time. The broadly investigated categories like sequential or session-centered recommendation are majorly used to suggest the next product based on a sequence of products and are less complex when compared to NCR. The recommendation and sequential modeling are carried out by NCR by taking into consideration a sequence of carts. It has been observed that to carry out sequential modeling, the recurrent neural network (RNN) has been evidenced to be extremely efficient and thus utilized for NCR. After carrying out an extensive analysis of the actual world datasets, it is observed that personalized product frequency (PPF) information supplies two significant signals for NCR. But it is claimed that in the recommendation scene, RNNs which exist are not capable of capturing information about the frequency of an item in NCR. PPF can be described as the number of times that every product is bought by an individual. However, the concept of PPF has been overlooked in related studies. As a consequence, prevailing approaches are not able to fully explore the significant signals which exist in PPF. Although prevailing approaches like RNN-centered techniques exhibit robust representation ability, the experimental studies demonstrate that these methods are not able to recognize PPF. By keeping in mind this constraint of RNNs, a straightforward product frequency-centered k-nearest neighbors approach is proposed to straightforwardly incorporate these significant signals. The four actual world datasets are used for the assessment of the proposed technique. Apart from the simplicity of the technique, the proposed method performs better than existing NCR techniques and other deep learning techniques which employ RNNs. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
Full text from SpringerLink