Large-Scale Item Categorization in e-Commerce Using Multiple Recurrent Neural Networks

Categories: E-commerce, Data Analysis | Pricing: Free | Official Website ↗

A deep learning method, DeepCN, uses multiple recurrent neural networks for precise item categorization in large-scale e-commerce.

This paper presents DeepCN (Deep Categorization Network), an end-to-end deep learning model designed for large-scale item categorization in e-commerce. It utilizes multiple recurrent neural networks (RNNs) to process various metadata attributes, generating features from textual data. These features are then fed into fully connected layers for classifying item categories. The model addresses challenges such as data size, category skewness, and noisy metadata by integrating diverse attributes into a common representation, which helps overcome sparsity and scalability issues. The categorization errors are back-propagated through the network to update weights in the RNNs during the learning process. DeepCN was evaluated on a large dataset of over 94 million items and approximately 4,100 leaf categories from a Korean e-commerce website, demonstrating improved accuracy compared to single RNN and unigram-based bag-of-words models.

Key Features

End-to-end deep learning model (DeepCN)
Utilizes multiple recurrent neural networks (RNNs)
Generates features from text metadata attributes
Classifies item categories using fully connected layers
Error back-propagation for weight updates
Integrates diverse attributes into common representation
Handles large-scale data and category skewness

Pros

Improves categorization accuracy over single RNN and bag-of-words models
Overcomes sparsity and scalability problems in e-commerce data
Integrates diverse metadata attributes effectively
End-to-end deep learning approach simplifies the process
Evaluated on a very large real-world dataset

Cons

This is a research paper, not a commercial product
Implementation details are for researchers, not end-users
No direct user interface or platform mentioned
Focuses on a specific e-commerce context (Korean website)
Requires deep learning expertise to implement

Use Cases

Automated item categorization in large e-commerce platforms
Improving search relevance through better product classification
Enhancing product discovery and recommendation systems
Managing large and noisy product metadata

Best For

E-commerce researchers
Data scientists in retail
Machine learning engineers
Academics studying deep learning applications

Watch demo on YouTube ↗

View full Large-Scale Item Categorization in e-Commerce Using Multiple Recurrent Neural Networks profile on Tools-Radar | Browse E-commerce tools | Alternatives to Large-Scale Item Categorization in e-Commerce Using Multiple Recurrent Neural Networks

Tools-Radar is a free directory of 10,000+ AI tools — discover, compare, and choose the right AI software for your needs. Visit tools-radar.com