在Airbnb的图形机器学习

Introduction

简介

Many real-world machine learning problems can be framed as graph problems. On online platforms, users often share assets (e.g. photos) and interact with each other (e.g. messages, bookings, reviews). These connections between users naturally form edges that can be used to create a graph.

许多现实世界中的机器学习问题都可以被框定为图问题。在网络平台上，用户经常分享资产（如照片）并相互交流（如信息、预订、评论）。用户之间的这些联系自然形成了可以用来创建图的边。

However, in many cases, machine learning practitioners do not leverage these connections when building machine learning models, and instead treat nodes (in this case, users) as completely independent entities. While this does simplify things, leaving out information around a node’s connections may reduce model performance by ignoring where this node is in the context of the overall graph.

然而，在许多情况下，机器学习从业者在建立机器学习模型时并不利用这些连接，而是将节点（在这里是指用户）视为完全独立的实体。虽然这确实简化了事情，但由于忽略了节点在整个图中的位置，忽略了节点的连接信息可能会降低模型性能。

In this blog post, we will explain the benefits of using graphs for machine learning, and show how leveraging graph information allows us to learn more about our users, in addition to building more contextual representations of them [4]. We will then cover specific graph machine learning methods, such as Graph Convolutional Networks, that are being used at Airbnb to improve upon existing machine learning models.

在这篇博文中，我们将解释使用图进行机器学习的好处，并展示如何利用图信息让我们了解更多关于用户的信息，此外还可以建立更多关于用户的上下文表示[4]。然后，我们将介绍具体的图机器学习方法，如图卷积网络，这些方法正在Airbnb使用，以改进现有的机器学习模型。

The motivating use-case for this work is to build machine learning models that protect our community from harm, but many of the points being made and systems being built are quite generic and could be applied to other tasks as well.

这项工作的动机是建立机器学习模型，以保护我们的社区免受伤害，但许多正在提出的观点和正在建立的系统是相当通用的，也可以应用于其他任务。

Challenges

挑战

When building trust & safety machine learning models around entities such as users or listings, we generally begin by reaching for features that directly describe the entity. For example, in the case of users, we may use features such as their location, account age, or n...