Mihajlo Grbovic’s research while affiliated with Yahoo and other places

What is this page?


This page lists works of an author who doesn't have a ResearchGate profile or hasn't added the works to their profile yet. It is automatically generated from public (personal) data to further our legitimate goal of comprehensive and accurate scientific recordkeeping. If you are this author and want this page removed, please let us know.

Publications (59)


TSMO 2024: Two-sided Marketplace Optimization
  • Conference Paper

August 2024

·

1 Read

Mihajlo Grbovic

·

Vladan Radosavljevic

·

Minmin Chen

·

[...]

·





Figure 1: Booking ranking position distribution.
Figure 2: Distributions of the suggested prices from different pricing strategy, and the actual price distribution.
reports few more statistics of our data set
Evaluation results of top 25 Markets using both data sets. We report the results for all pricing models, and baselines. Our REV_POTENT is better than CRM and baselines. Our BR is almost zero, with a RECALL almost 1.0. PDR PDR_HP PDR_LP PDR_HP/PDR_LP
Revenue Maximization of Airbnb Marketplace using Search Results
  • Preprint
  • File available

November 2019

·

110 Reads

Correctly pricing products or services in an online marketplace presents a challenging problem and one of the critical factors for the success of the business. When users are looking to buy an item they typically search for it. Query relevance models are used at this stage to retrieve and rank the items on the search page from most relevant to least relevant. The presented items are naturally "competing" against each other for user purchases. We provide a practical two-stage model to price this set of retrieved items for which distributions of their values are learned. The initial output of the pricing strategy is a price vector for the top displayed items in one search event. We later aggregate these results over searches to provide the supplier with the optimal price for each item. We applied our solution to large-scale search data obtained from Airbnb Experiences marketplace. Offline evaluation results show that our strategy improves upon baseline pricing strategies on key metrics by at least +20% in terms of booking regret and +55% in terms of revenue potential.

Download

Real-time Personalization using Embeddings for Search Ranking at Airbnb

July 2018

·

1,628 Reads

·

313 Citations

Search Ranking and Recommendations are fundamental problems of crucial interest to major Internet companies, including web search engines, content publishing websites and marketplaces. However, despite sharing some common characteristics a one-size-fits-all solution does not exist in this space. Given a large difference in content that needs to be ranked, personalized and recommended, each marketplace has a somewhat unique challenge. Correspondingly, at Airbnb, a short-term rental marketplace, search and recommendation problems are quite unique, being a two-sided marketplace in which one needs to optimize for host and guest preferences, in a world where a user rarely consumes the same item twice and one listing can accept only one guest for a certain set of dates. In this paper we describe Listing and User Embedding techniques we developed and deployed for purposes of Real-time Personalization in Search Ranking and Similar Listing Recommendations, two channels that drive 99% of conversions. The embedding models were specifically tailored for Airbnb marketplace, and are able to capture guest's short-term and long-term interests, delivering effective home listing recommendations. We conducted rigorous offline testing of the embedding models, followed by successful online tests before fully deploying them into production.


Figure 1: Toy example of user session log with purchase intent.
Figure 2: Graphical representation of Deep Memory Network for predicting mobile purchases based on observed actions
Figure 3: Recall@K for DMN vs. DMN-user model
Recall@5 for DMN model over baselines on purchase items prediction task.
Hand curated 10 app events among 50 app events in the embedded space for query app events Snapchat: open and Snapchat:close as learned by the DMN model.
Modeling Mobile User Actions for Purchase Recommendation using Deep Memory Networks

June 2018

·

164 Reads

·

16 Citations

Rapid expansion of mobile devices has brought an unprecedented opportunity for mobile operators and content publishers to reach many users at any point in time. Understanding usage patterns of mobile applications (apps) is an integral task that precedes advertising efforts of providing relevant recommendations to users. However, this task can be very arduous due to the unstructured nature of app data, with sparseness in available information. This study proposes a novel approach to learn representations of mobile user actions using Deep Memory Networks. We validate the proposed approach on millions of app usage sessions built from large scale feeds of mobile app events and mobile purchase receipts. The empirical study demonstrates that the proposed approach performed better compared to several competitive baselines in terms of recommendation precision quality. To the best of our knowledge this is the first study analyzing app usage patterns for purchase recommendation.


Workshop on Two-sided Marketplace Optimization: Search,Pricing, Matching & Growth

February 2018

·

65 Reads

·

1 Citation

The 1st International Workshop on Two-sided Marketplace Optimization: Search, Pricing, Matching & Growth(TSMO) will be held in Los Angeles, California, USA on February 9th, 2018, co-located with the 11th ACM International Conference on Web Search and Data Mining(WSDM). The main objective of the workshop is to address the challenges of two-sided marketplace optimization in web-scale settings. The workshop brings together interdisciplinary researchers in information retrieval, recommender systems, personalization, and related areas, to share, exchange, learn, and develop preliminary results, new concepts, ideas, principles, and methodologies on applying data mining technologies to marketplace optimization. We have constructed an exciting program papers and invited talks that will help us better understand the future of two-sided marketplaces


Search Ranking And Personalization at Airbnb

August 2017

·

1,229 Reads

·

14 Citations

Search ranking is a fundamental problem of crucial interest to major Internet companies, including web search engines, content publishing websites and marketplaces. However, despite sharing some common characteristics a one-size-fits-all solution does not exist in this space. Given a large difference in content that needs to be ranked and the parties affected by ranking, each search ranking problem is somewhat specific. Correspondingly, search ranking at Airbnb is quite unique, being a two-sided marketplace in which one needs to optimize for host and guest preferences, in a world where a user rarely consumes the same item twice and one listing can accept only one guest for a certain set of dates. In this talk, I will discuss challenges we have encountered and Machine Learning solutions we have developed for listing ranking at Airbnb. Specifically, the listing ranking problem boils down to prioritizing listings that are appealing to the guest but at the same time demoting listings that would likely reject the guest, which is not easily solvable using basic matrix completion or a straightforward linear model. I will shed the light on how we jointly optimize the two objectives by leveraging listing quality, location relevance, reviews, host response time as well as guest and host preferences and past booking history. Finally, we will talk about our recent work on using neural network models to train listing and query embeddings for purposes of enhancing search personalization, broad search and type-ahead suggestions, which are core concepts in any modern search.



Citations (43)


... Algorithms, metrics, and models for carousel recommendations At the same time, numerous commercial system providers have demonstrated the positive effects of carousel-based interfaces. The authors of the corresponding publications have proposed algorithmic improvements, e.g., to optimize how the collections are ordered among each other, how they are filled with items, and how labels are assigned (Wu et al., 2016(Wu et al., , 2021McInerney et al., 2018;Bendada et al., 2020;Lo et al., 2021). Singal et al. (2021) even investigated how to implement carousels independently of the underlying algorithms, requiring only standard user-item representations. ...

Reference:

Multi-List Interfaces for Recommender Systems: Survey and Future Directions
Toward User Engagement Optimization in 2D Presentation
  • Citing Conference Paper
  • March 2021

... Determining the population that should take part in the A/B test. Examples include a simple 50/50 split of all users [163], an assignment where the target population is determined over a two week period [173], or an assignment where network effects have to be taken into account [102]. • Determining the goal or hypothesis for A/B testing is frequently mentioned for multi-armed A/B tests (17 occurrences). ...

How Airbnb Tells You Will Enjoy Sunset Sailing in Barcelona? Recommendation in a Two-Sided Travel Marketplace
  • Citing Conference Paper
  • July 2020

... General Product Rec +1.58% +1.49% +0.33%The third type are graph learning methods.• GroupID[12]: a meta-learning approach proposed by Airbnb to generate new item embeddings by calculating the average of similar item embeddings. • GIFT[21]: a meta-learning approach to address the coldstart problem by generating ID embeddings for cold items using other available features.5.1.3 ...

Real-time Personalization using Embeddings for Search Ranking at Airbnb
  • Citing Conference Paper
  • July 2018

... Despite the existing success, designing and selecting appropriate features for different tasks remain a very challenging problem. 2 Significant advances with the aforementioned challenges have been attained more recently, with the ef-forts in developing deep representations of activities to help automatically learn their features. 3,4 Even though these methods model sequences of activities and apply different strategies to tackle or filter noisy data such as various attention mechanisms, there are several challenges remaining largely untackled in the practice. We highlight two emerging ones in this study. ...

Modeling Mobile User Actions for Purchase Recommendation using Deep Memory Networks

... Instead, we start to accept work as a feature and a product of the platform (see Kässi and Lehdonvirta, 2018). We turn to the technology as a mode of affordance and to its owners as those who are responsible to make it work for us (see Allio, 2004;Benson et al., 2017;Calo and Rosenblat, 2017;Dellarocas, 2010;see Dellarocas et al., 2006;Evans and Schmalensee, 2016;Grbovic and Noulas, 2018;Lee et al., 2018;McAfee and Brynjolfsson, 2017;Rosenblat, 2018;Schmitz, 2018;Warhurst et al., 2017). ...

Workshop on Two-sided Marketplace Optimization: Search,Pricing, Matching & Growth
  • Citing Conference Paper
  • February 2018

... Pooled ridesharing represents a significant evolution in the transportation landscape, offering a model where riders share trips with others, potentially reducing the number of vehicles on the road and fostering environmental benefits. Uber Pool and Lyft Shared, introduced in 2014, popularized this concept, achieving initial success in urban markets such as San Francisco, Chicago, and New York City [29][30][31][32]. PR offers an efficient solution for reducing traffic congestion by consolidating multiple passengers into a single ride. ...

Analyzing Uber's Ride-sharing Economy
  • Citing Conference Paper
  • April 2017

... They combined all of the features into a Maximum Entropy model [64] for app categorization. Radosavljevic et al. [65] took advantage of users' app installation behavior to categorize apps. He et al. [66] explored sequential characteristics of app usage. ...

Smartphone App Categorization for Interest Targeting in Advertising Marketplace
  • Citing Conference Paper
  • April 2016

... Additionally, search and recommender systems have increasingly represented model inputs as categorical features, shifting away from the focus on featurizing these inputs using more traditional and usecase specific methods. For example, queries (Grbovic et al., 2015), documents (Le et al., 2018), metadata (Vasile et al., 2016), and users have all been treated as categorical features in modern search and recommendation systems. This strategy has proven successful, improving model performance and user experiences. ...

Search Retargeting using Directed Query Embeddings
  • Citing Conference Paper
  • May 2015