Improving product discovery via relevance and ranking optimization
Submitted by Akash Khandelwal (@akash099) on Saturday, 31 March 2018
In e-commerce, recommendations play a key role not only in customer satisfaction by improving discovery but also helps fulfill business objectives. In this talk, I will focus on our iterative journey starting from feature engineering, adding features incrementally and learning on them, thus moving from a rule based system to launching a machine learnt system in production.
The Flipkart product recommender system consist of two layers - relevance and ranking.
1. Relevance layer takes into account user behavioural events including browse and purchases and computes similarity score for product pairs using collaborative filtering. This similarity score is used to generate a selection set for powering product recommendations.
2. In the Ranking Layer, we rank these selected candidates using Learning to Rank technique to generate our final recommendation list. As a part of this talk, I will cover the feature set and algorithms used for predicting the conversion probability of recommended products.
The talk will cover following topics :
a) User shopping journey at Flipkart and importance of product discovery
b) Types of product recommendations : similar products, cross selling etc.
c) Architecture of Recommender System : Relevance and Ranking modules
d) Using product textual and visual attributes for computing product similarity
e) Using crowdsourced activity data to compute the set of relevant products
f) Formulation of ranking as an machine learning problem towards optimising conversion rates
g) Our learnings from various iterations over feature-sets and ML models
Akash is a software developer with Search Relevance team at Flipkart, working on improving Autosuggest. Previously, he has worked on building Flipkart Recommendation System. He designed real time and batch pipelines to power recommendations, including use cases such as product bundling, similar products and personalisation. He is interested in applying Machine Learning for pattern mining, and deploying data processing pipelines at scale. He graduated with a dual degree in Computer Science & Engineering from IIT Delhi.