The project team aims to train a model to predict whether or not a user will like a particular song.
Three things I like:
1) The tool can be pretty helpful. Obviously, it can save people a lot of time if they do not have to run through a whole list of songs only to realize that they don't like most of them.
2) The dataset contains a lot of features, and I believe they will help train a reliable model.
3) It is good that you can extract data from Spotify through an API. Otherwise it will be pretty painstaking to obtain so much information about songs.
Three concerns:
1) It seems that a model will be user-specific. If you provide a model for Vidita in your final report, I will not be able to use the model to select the songs I like unless you train another model for me.
2) The y matrix will be very subjective. People press "like" or "dislike" based on many subjective and even random factors, not necessarily predictable, and even not necessarily related to the songs themselves... Even if you are able to get a model, I will doubt its validity.
3) You may be incorporating too many ideas. It would be nice for you to focus on question for now.
The project team aims to train a model to predict whether or not a user will like a particular song.
Three things I like: 1) The tool can be pretty helpful. Obviously, it can save people a lot of time if they do not have to run through a whole list of songs only to realize that they don't like most of them. 2) The dataset contains a lot of features, and I believe they will help train a reliable model. 3) It is good that you can extract data from Spotify through an API. Otherwise it will be pretty painstaking to obtain so much information about songs.
Three concerns: 1) It seems that a model will be user-specific. If you provide a model for Vidita in your final report, I will not be able to use the model to select the songs I like unless you train another model for me. 2) The y matrix will be very subjective. People press "like" or "dislike" based on many subjective and even random factors, not necessarily predictable, and even not necessarily related to the songs themselves... Even if you are able to get a model, I will doubt its validity. 3) You may be incorporating too many ideas. It would be nice for you to focus on question for now.