Product categorization with Machine Learning
MercadoLibre is the largest online retailer in Latin America with over 100M users. It is an e-commerce and auction platform that provides users with buying and selling mechanisms for e-transactions. We established a working relationship with them in order to solve issues related to the category tree of their product catalog.
About MercadoLibre
Headquarters: Buenos Aires, Argentina
Founded: 1999
Stock: NASDAQ, MELI
Efficiency and automation boost
Categorization has become much more efficient and we implemented some automation of category creation and product classification.
Approach
The problem was faced in three stages: Research & Development, bringing the model into production, and exposing it through an API and Frontend development.
Stage 1
Data analysis
R&D implied a massive historical data analysis which later directed us to the hypothesis, trial & error circle until a fair solution was found.
Stage 2
Machine Learning modeling
We built an algorithm that uses neural networks and users' navigation data to facilitate item categorization.
We also developed a set of algorithms using deep learning and NLP tools to suggest missing categories automatically. Some of the tools used where Gensim, Keras, and NLTK.
Technical details
Stage 3
Frontend Solution
As human validation is part of the solution, we built an ultra-intuitive and intelligent frontend using React + Redux, ES6, and Webpack to communicate with the built API.
Testimonial
Tryolabs has been a key piece in optimizing and extending our category trees at a much higher pace than we could achieve with our former tools and teams. Through the mastering and customization of state of the art Machine Learning techniques, we've been able to increase item "findability" and conversion rates. We are so glad to have them collaborating with us in the quest for the best e-commerce solution. I would absolutely recommend them.
Diego Cabrera
SENIOR MANAGER, MERCADO LIBRE
Get in touch
Tell us a bit about your project, we'd love to see how we can help.
You can send us an email or just fill out the form below.
Email us!
