Abstract: In machine learning, transfer learning has become a potent technique that enables models to use their previous knowledge and representations to perform better on new tasks or domains. By using pre-trained models that have acquired general features from massive datasets or comparable jobs, this strategy gets beyond the constraints of data availability and processing resources. several approaches to knowledge transfer between domains are provided by several forms of transfer learning, including inductive, transductive, unsupervised, semi-supervised, multi-task, and zero-shot transfer learning. Transfer learning has many uses in e-commerce, including fraud detection, product classification, and recommendation and picture search. This paper gives a thorough guide to classifying e-commerce products using transfer learning approaches, highlighting the advantages of knowledge transfer and the efficiency of fine-tuning models for particular use cases.

Keywords: transfer learning, e-commerce, image classification, text classification, data collection, fine tuning.

| DOI: 10.17148/IARJSET.2023.107124