Data Science Battleground #8: Let’s create your own movie recommender system

Last week on May 11th, Tokyo Techies welcomed 10 participants joining with us in our intensive but interesting Data Science Battleground Round 8: What to Watch – How a recommender system works.

Like in other previous rounds of Data Science Battleground, our participants are empowered to acquire fundamental Data Science knowledge as well as topic-related Data Science skills through high-quality lectures prepared by our mentors. What’s more, they can practice all the skills and code they have learned right after the lecture by teaming up to compete against each other in our special battleground. We will select the winner based on the final evaluation score. Different topics will have different evaluation style.

Although recommender system is a challenging topic in Data Science, we were glad to welcome10 participants joining with us in this round. 

What is a recommender system?

The concept of filtering and personalizing data has always been attractive, especially in this new era, when data is the new oil. The internet is like the Wild West, with all The Good, The Bad and The Ugly. Is there anything we could do to utilize all these data? There has to be a way to do that. That is why the recommender system comes to life.

Most internet based products and contents these days are powered by the recommender system. Facebook, Netflix, Youtube, Amazon and thousands of similar websites utilized recommender system to filter out millions gigabyte of data and make those data personalized to users. With the help of the recommender system, tremendous values were provided to the internet businesses and their consumers. Over the years, the recommender system algorithms have been improved by tenfold to deliver personalized content to the consumers. Even though the recommender system sounds like the secret sauce to every multi-millionaire business, it is not an easy task to build a successful recommender system model.

What have our participants experienced in our round 8?

Our topic this time is a movie recommender system. The concept might sound complicated, however, each of the participants was able to build their own personalized recommender yourself. Although there are different approaches to build a recommender system model, in the last event on May 11th, we use collaborative filtering for our model. It builds the model from the user’s past behaviors such as previously interacted products or the ratings our participants have given to those items. The model then used those data to predict movies that the user might have an interest in.

In order to make sure that our audience, some of whom are totally beginner in Data Science field of expertise, was all on the same page, mentor Hiep Nguyen carefully instructed them on the recommender system and collaborative filtering, while mentor Nghia Truong explained to all participants about the movie recommender system model that they would work on later in the teamwork battle session. They received and answered several provoking questions from the audience. Due to this interactive lesson, our participants found it easier for them not only to acquire fundamental knowledge about the system but also to put them into practice for the next session – “battleground” of the workshop.

Mentor Hiep Nguyen instructed participants on the foundation knowledge about the recommender system.

Mentor Nghia Truong introduced the movie recommender system model

In this battleground,  each team spent one hour tweaking the codes to improve the pre-written recommender system model for a more accurate movie recommendation list by finding the best numbers for our RMSE evaluations. This battleground was said to be impressively effective as our participants can both review what they have learned and apply those immediately into practice. We were so glad that our participants discussed with each other and shared their knowledge from various background to come up with the best solution for recommender system model improvement. Thanks to their active participation, in the end, we had a wonderful battleground which empowered our participants to experience hands-on practice and to make friends with several interesting people!

If you are yearning for more knowledge and new friends, let’s join the Data Science Battleground round 9: Costa Rican Household Poverty Level on May 25th, 2019. Let’s learn, compete and make friends!

© 2020 Tokyo Techies. All rights reserved.