emotion-classification

Arousal and Valence Prediction Model

This project aims to predict the arousal and valence levels of songs based on extracted features from audio data. A Random Forest Regressor model is trained using statistical features of the song's arousal and valence data to predict the emotional tone of the song.

Steps in the Code

1. Data Loading

• The code loads CSV files containing the arousal and valence features for each song. These files are assumed to contain data about songs over different time points.

2. Feature Extraction

• A function extract_summary_statistics calculates summary statistics (mean, median, standard deviation, max, and min) for each song's arousal and valence data. These statistics summarize the emotional tone of the song.

3. Data Preparation

• The summary statistics (features) for both arousal and valence are combined into a single matrix X, which will be used as the input for training the models.
• The labels (y_arousal and y_valence) are created by calculating the mean of the arousal and valence values across all time points.

4. Data Splitting

• The dataset is split into training and testing sets, with 80% used for training and 20% for testing.

5. Model Training

• Two Random Forest Regressor models are trained:
  • One model for arousal prediction.
  • One model for valence prediction.
• These models learn to predict the emotional tone of the song based on the extracted features.

6. Model Evaluation

• The models are evaluated using Mean Squared Error (MSE) to measure the accuracy of the predictions. A lower MSE indicates better performance.

7. Saving the Models

• The trained models are saved using joblib, allowing them to be reused later without needing to retrain them.

8. Output

• The MSE values for both models are printed to show how well the models are performing on the test data.

Requirements

• Python 3.x
• Libraries:
  • numpy
  • pandas
  • scikit-learn
  • joblib

How to Run

1. Ensure you have the required libraries installed:

   pip install numpy pandas scikit-learn joblib
   

2. Make sure the arousal.csv and valence.csv files are in the correct directory (./data/annotations/).
3. Run the script:

    python model_training.py