A new model for assessing the role of customer behavior history, product classification, and prices on the success of the recommender systems in e-commerce

Wakil, Karzan; Alyari, Fatemeh; Ghasvari, Mahdi; Lesani, Zahra; Rajabion, Lila

doi:10.1108/k-03-2019-0199

Cited by 23 publications

(24 citation statements)

References 93 publications

(69 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Baselines are also pretrained on the Taobao data and initialized following their original papers. We pretrain two types of models for baselines: (1) training with normal image and text modality which include "ViLBERT", "LXMERT" and "VLBERT"; (2) training with image, text, and knowledge modality which include "ViLBERT+PKG", "VLBERT+PKG", and "LXMERT+PKG", where knowledge text from PKG are spliced behind title text as the text modality input. More details are in Appendix.…”

Section: Pretraining Of Baselinesmentioning

confidence: 99%

“…The emergence of E-commerce has greatly facilitated people's lives. And there are a large number of product-based application tasks in the E-commerce scenario, such as item classification [1,2], product alignment [3], recommendation system [2,4,5] and so on. As shown in Figure 1, there are usually images, titles, and structure knowledge of products, which is a typical multi-modal scenario.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Knowledge Perceived Multi-modal Pretraining in E-commerce

Zhu

Zhao

Zhang

et al. 2021

Proceedings of the 29th ACM International Conference on Multimedia

View full text Add to dashboard Cite

In this paper, we address multi-modal pretraining of product data in the field of E-commerce. Current multi-modal pretraining methods proposed for image and text modalities lack robustness in the face of modality-missing and modality-noise, which are two pervasive problems of multi-modal product data in real E-commerce scenarios. To this end, we propose a novel method, K3M, which introduces knowledge modality in multi-modal pretraining to correct the noise and supplement the missing of image and text modalities. The modal-encoding layer extracts the features of each modality. The modal-interaction layer is capable of effectively modeling the interaction of multiple modalities, where an initial-interactive feature fusion model is designed to maintain the independence of image modality and text modality, and a structure aggregation module is designed to fuse the information of image, text, and knowledge modalities. We pretrain K3M with three pretraining tasks, including masked object modeling (MOM), masked language modeling (MLM), and link prediction modeling (LPM). Experimental results on a real-world E-commerce dataset and a series of product-based downstream tasks demonstrate that K3M achieves significant improvements in performances than the baseline and state-of-the-art methods when modality-noise or modality-missing exists.

show abstract

Section: Pretraining Of Baselinesmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Knowledge Perceived Multi-modal Pretraining in E-commerce

Zhu

Zhao

Zhang

et al. 2021

Proceedings of the 29th ACM International Conference on Multimedia

View full text Add to dashboard Cite

show abstract

“…In essence, a weighted vote is nothing more than the weight obtained by the When a correct prediction is produced, the confidence of the base classifier is calculated, which is generally computed by the related meta classifier when a successful prediction is made. Moreover, in his work, Yu et al [17] proposes grading as a generalization method for cross-validation selection, in which the training set is divided into n subsets and an n-1 classifier is constructed, therefore removing one split at a time to investigate the speed of misclassification [24]. Finally, this strategy chooses the learning classifier with the lowest rate of misclassification.…”

Section: Gradingmentioning

confidence: 99%

An Ensemble and Filtering-Based System for Predicting Educational Data Mining

Hananto¹

2021

J. Appl. Data Sci.

View full text Add to dashboard Cite

When developing a prediction paradigm, an ensemble technique such as boosting is used. It is built on a heuristic framework. Generally speaking, engineering ensemble learning is more accurate than individual classifiers when it comes to making predictions. Consequently, numerous ensemble strategies have been presented in this work, particularly to provide a more complete understanding of the essential methods in general. Researchers have experimented with boosting methods to forecast student performance as part of a variety of ensemble techniques. The researchers employed improvement approaches to construct an accurate predictive educational model, which was based on a key phenomena seen in categorization and prediction operations. In light of the uniqueness and originality of the suggested strategy in educational data mining, the researchers used augmentation strategies in order to construct an accurate predictive pedagogical model. Tenfold cross-validation was performed to evaluate the effectiveness of the basic classifiers, which included the random tree, the j48, the knn, and the Naive Bayes. The random tree was found to be the most effective classifier. Several additional screening techniques, including oversampling (SMOTE) and undersampling (Spread subsampling), were utilized to analyze any statistically significant variations in results between the meta and base classifiers that had been identified between the meta and base classifiers. The use of ensemble and screening strategies, as compared to the use of standard classifiers, has demonstrated considerable gains in predicting student performance, as has the use of either strategy alone. Furthermore, after the completion of a performance research on each approach, two new prediction models have been established on the basis of the improved results gained thus far.

show abstract

“…the entries with business field of zero, were processed according to the business logic. First, the entries of the stores with no transaction were deleted; Second, the outliers containing the information of actual transactions were retained; Third, the other outliers that deviate greatly from the normal range were processed to ensure the accuracy of subsequent normalization [17].…”

Section: Data Processing and Index Selectionmentioning

confidence: 99%

A Hierarchical Model of E-Commerce Sellers Based on Data Mining

Bai¹

2020

ISI

View full text Add to dashboard Cite

This paper attempts to accurately classify e-commerce sellers based on data mining. Firstly, the original data from an e-commerce platform were preprocessed, and the classification indices were identified from five categories (products, users, traffic, sales and basic attribute). Next, the principal component analysis (PCA) and the self-organizing feature map (SOM) were fused into a hierarchical model that divides e-commerce sellers into three categories: large sellers, medium sellers and small sellers. The effectiveness of our model was verified through experiments. Finally, several operating strategies were put forward for e-commerce sellers in each category. The research results provide a good reference for the development of the e-commerce industry.

show abstract

A new model for assessing the role of customer behavior history, product classification, and prices on the success of the recommender systems in e-commerce

Cited by 23 publications

References 93 publications

Knowledge Perceived Multi-modal Pretraining in E-commerce

Knowledge Perceived Multi-modal Pretraining in E-commerce

An Ensemble and Filtering-Based System for Predicting Educational Data Mining

A Hierarchical Model of E-Commerce Sellers Based on Data Mining

Contact Info

Product

Resources

About