HitHub — a hit list predictor #Week 6
Hi! The previous week, we said that we will divide the music by region, but we gave up on that because we had to take advantage of a different list for each country and our team couldn’t be sure whether they were reliable or not. Fortunately, we had one more idea, which is to use the Billboard Global 200 list.
Billboard Global 200 chart is also coming from the company Billboard, but it does not just count the sales and streamings in just the United States. According to Wikipedia, it counts the sales and streamings from over 200 territories worldwide.
But there is an issue with this list. It is launched in September 2020. The list is just 9 months old. The problem is, we could not have a good amount of data, contrary to the Billboard Hot 100, which is launched in 1955.
As a result, we changed our minds again and decided to stick with the Billboard Hot 100. Eventually, the model is going to be output whether the song is eligible to enter Billboard Hot 100 list.
In addition, we have not decided which model is going to perform better. Hence, we are preparing something so that we can test all possible situations and give us the results so that we can choose the model that gives the most accurate result and work on it. As you know, we have Logistic Regression, KNN, SVM, Gaussian Naive Bayes, and Artificial Neural Network (ANN). For example, we aim to select the important hyperparameters on ANN and run them in different combinations and get the most accurate results.
We are finalizing our project and looking forward to getting the results that we want. 🎧🎧🎧
