Discover Your Music Style with Data
Have you ever wondered if your hobbies and social background can predict your taste in music? In a recent project, we explored just that! By analyzing data from a large dating service, we created a tool that uses demographic and hobby data to predict music preferences. Here’s how we did it.
The Dataset
We used anonymized data from around 120,000 members of a European dating service. The data included 24 attributes, such as age, education level, and hobbies, as well as information on their musical preferences. Our goal was to understand how these variables influence people’s tastes in music.
Our Approach
To create our model, we built a self-organizing map (SOM), which is a type of machine learning model used to visualize complex data. In this case, the SOM helped us group people based on their social and hobby-related attributes, making it possible to predict their musical tastes.
Here’s how it works: by answering a few simple questions about yourself (like your age and hobbies), you can find out which group you belong to in the map. Then, based on your group, we predict which music styles you're most likely to enjoy. The model takes into account 8 different genres, from Rock/Pop to Jazz/Blues.
What Can You Learn?
By testing your own profile against the model, you can see where you fall compared to other respondents. For each music genre, the model gives you a score, showing how your preferences match up with others in your group. Maybe you’ll discover you’re more into Jazz than you thought, or perhaps your love for Rock is right on point!
Why It’s Useful
Understanding personal preferences, especially when only limited data is available, can be tricky. However, models like this have broad applications. They can be used in product design, marketing, customer engagement, and even social network analysis. This kind of data-driven insight helps tailor products and services to specific target groups.
How We Built It
We preprocessed the data by standardizing variables and ensuring no missing information. Then, we used the SOM model to group respondents into approximately 400 clusters, allowing us to create detailed predictions for each person’s music preferences.
The results were based on attributes such as age, education level, and hobbies like theater, literature, and dance. After refining the model, we ended up with a map that can help predict music styles based on someone’s profile.
Try It Yourself
Curious to see how it works? In our showcase, you can input your own social and hobby preferences to find out which music styles people like you tend to enjoy. It’s a fun way to explore how your interests might influence your taste in music