Kaggle Competition - The Perfect Tune

I used R to construct supervised machine learning models with a dataset of popular songs, to predict Spotify ratings based on auditory features of the songs.

Project Description:

• This is a Kaggle competition I participated in November 2022.
• It provides dataset that describes popular songs based on auditory features, such as loudness, tempo, performer and genre.
• The goal is to contruct a predictive model using analysisData.csv, and predict the songs' ratings based on the features in scoringData.csv.
• Model performance will be evaluated based on RMSE (root mean squared error).

Data Analysis Process:

Please view the detailed raw R codes and coding results for the entire data analysis process as an illustration.

1. Data Exploration
2. Data Tidying
   • Encode Missing Data
   • Data Parsing
   • Data Transformation
3. Feature Selection
   • Variable Inter-set
   • Remove Near Zero Variance
   • Principal Components Analysis
   • Split Data
4. Data Analysis - Modeling
   • Multiple Regression
   • Regression Tree
   • Random Forest
   • Ranger
   • XGBoost
   • gbm
   • Radial SVM
   • Model Tuning
   • Comparison
5. Prediction