love-interest-spotify

Over the years, I have had the opportunity to encounter multiple failed attempts at queer romance. These romantic encounters vary from ex-boyfriends in long-distance relationships, delusionships i.e. having an requited crush on someone, and the person lives in my head rent-free, situationships i.e. either my romantic partner or I (the "or" is inclusive) want the fun of a relationship but denies the responsibilities of the actual label, and the list goes on. I believed that it was a completely normal and healthy behavior to give each of these love interests a one to two-hour long Spotify playlist that is purely dedicated to my feelings for them and our experiences between each other. While the class of love interest I have varies in its categories and thrives in diversity, due to time and mental constraints, I wish to explore a machine learning model that can classify whether a track is about my latest situationship, whose arbitrary name is James. The choice for this arbitrary name is due to the fact that this love interest is British. My exploratory challenge is as follows: "Can a logistic regression model classify whether this song is about James?" This report explores the classification model using audio features and metadata of the Spotify tracks as predictors in order to produce a classification. The report justifies the choice of predictors and processes the data before training the model. After model training, the model evaluates the trained model based on model performance metrics, including precision, accuracy, recall and F-1 score, before finding the potential best parameters for the model.