Title: 150 Successful Machine Learning Models: 6 Lessons Learned at Booking.com
Rating: 4/5
Summary: The article explains six important lessons that the authors experience while de-veloping 150 machine learning models at Booking.com. They explain each lesson with real world
examples and demonstrates it’s important to consider the business, analyze the results, and im-prove the models. The article is helpful for people who work with machine learning in business and for researchers research it.
Strengths:
1. The paper demonstrates six important things the authors learned from developing and deploying machine learning models at Booking.com. The authors explain each lesson with real examples, so it helps me understand and motivated me to use it for my own projects.
2. Figure 4 shows that the fact that a model is improved, doesn’t always mean more people will purchase something at Booking.com. Therefore, when people try to adopt new technologies
in the real world business, they have to think about many things related to real business. The study done in constrained environments may not be true in the real world.
3. Section 7 talks about the importance of conducting randomized controlled trials to test the impact of models on businesses and provides ideas on how to design experiments for that purpose. It was amazing to see how they used their creative thinking to simulate real-life situations, run experiments, and make use of them for business.
Weaknesses
The paper may not be useful for all businesses because it talks only about how Booking.com used machine learning models. While the authors provide general advices, readers need to adapt it and customize it.
Questions:
1. The authors explain how important it is to know about the business and collect the data for machine learning models. Can they be sure they have right and enought data? Do they consider about privacy while they collect data?
2. The paper explains how important it is to be able to understand how machine learning models work in the author’s setting. But I wonder what kind of concrete things I can do to make sure I can understand these models?
Discussions
After I read the paper, I learned more about how to make machine learning models for businesses. The paper says it’s important to understand what businesses and users need and keep making models better. As a researcher, I want to use these tips in my work to make models that fit businesses and users. Also, the paper makes me want to find more ways to make machine learning models easier to understand.