3.3.3. Light Gradient Boosting Machine (LightGBM)

Light Gradient Boosting Machine is a method that builds multiple decision trees in a way that is faster and more efficient than traditional methods . Traditional decision tree methods grow the tree level-by-level, meaning they expand each branch across all nodes before moving deeper into the tree. However, LightGBM, instead of expanding every part of the tree evenly, it focuses on growing branches only where they can make the biggest improvement to the model and eventually finds the leaf with highest potential and splits that one. This makes LightGBM be able to handle big datasets more quickly without losing accuracy. This approach could be helpful when we need to balance speed with precision, such as in emotion recognition tasks that proces lots of sensors and signal data.

By merging both DREAMER and DERCFF datasets, the following two studies were able to achieve significant results. In a study by Dua , LightGBM was also used to analyze ECG signals and facial features, achieving a high accuracy of 92.7%, demonstrating high performance. Another study, EmoSens by G. S et al., applied LightGBM and achieved a 92.5% recognition rate for nine emotion categories .