Slideshow by Diana Herrera-Gonzalez

What is machine learning?

dianaherreragonzale9 — 2025-05-15 16:29:03 UTC

Machine learning is the use and development of computer systems that are able to learn and adapt without following any directions. Things such as predictions and decisions are all from their instincts.

"A baby learns to crawl, walk, and run. We are in the crawling stage when it comes to applying machine learning." -Dave Waters

Each filtering method

dianaherreragonzale9 — 2025-05-15 18:09:38 UTC

When you're rating your 23 songs, you start off with content-based filtering, you're listening to these randomized songs based on tempo, lyrics, genre etc. Once you rate them based on your opinion, it ranks new songs with similar features, it moves onto collaborative filtering. Collaborative filtering is when they add in more new songs and rate it based on the data that you've added when rating those 23 songs. In apps like Spotify, Pandora, or other music platforms use this kind of filtering to add in more music that suits your own preferences based on your liked songs, the artists you follow on the platform and your own playlists.

Why we rate 23 songs?

dianaherreragonzale9 — 2025-05-15 18:25:39 UTC

The reasoning for rating twenty three songs was because when you rate these twenty three songs, you're giving the Colab more data of your own music taste. So when the Colab gives you different songs, they'll use the data from those twenty three different songs as a reference and rate based on your own interests. This represents machine learning by using your ratings on the songs, colab builds a model to learn patterns of your preferences, make predictions based on the new songs, and lastly improves as you rate more songs.

Results

dianaherreragonzale9 — 2025-05-20 18:35:55 UTC

When it came to my actual results, I was surprised by the amount of work put into the coding and how accurate it could be at times. This project overall was teaching my class and I how data can be interpreted into apps that we use daily. For example, on Spotify they use our liked songs, the artists we follow and our playlists to recommend us new music and much more, this overall taught me how the coding works and the entire process of collaborative filtering.

attribute model

dianaherreragonzale9 — 2025-05-20 18:37:20 UTC

One main attribute model of mine was tempo, it was more suited for me based on my ratings, when it rated another song based on my preference it was more accurate than I expected it'd be.

Common questions

dianaherreragonzale9 — 2025-05-20 18:41:50 UTC

My results compared to the actual rating were at times very close, and sometimes were very far off. Some surprises where how accurate the guesses were, they knew how to determine my ratings based on the twenty three songs I had rated based on my personal preference. This adds to my understanding of collaborative filtering by the code recommending me songs based off my own interest and also how they can send in my data into ratings of other songs. Some changes that could happen would be my own personal preference days later, and this could also have the code change my ratings in the same songs or different songs.

My thoughts

dianaherreragonzale9 — 2025-05-20 18:42:41 UTC

I personally found this project very interesting because I have never given thought on what kind of data and coding is put into things such as suggesting songs based on your own ratings. Throughout this project I learned that many things need to be incorporated into suggestions in music, movies, shows etc. I also learned what machine learning, and collaborative filtering was all about, and that was learning the data and making their own conclusions based on the data provided.

class rating

dianaherreragonzale9 — 2025-05-26 05:02:37 UTC

When it came to my class ratings, it was closely similar to my own rankings because even with little data, their can always be a similar prediction for everyone.

attribute model 2

dianaherreragonzale9 — 2025-05-26 05:04:16 UTC

Just like my attribute model one, it was just as accurate, I was very impressed that even with such a small amount of data it would be able to predict my own music preferences based on my attribute which was unique words.