Mouse Aggression Classification: 8-Model Comparison 7-CLASS

Aggression Behavior Classification (Classes 1-7, Excluding Class 0) • 5 Independent Runs with Different Random Seeds

Experiment Configuration Task: 7-class aggression behavior classification | Data: 67,039 samples × 7 features | Split: Stratified 70/15/15 with seeds 42-46 | Metrics: Mean ± SD across 5 runs

📊 Key Observations (7-Class Aggression)

XGBoost achieves the best performance (63.2% accuracy), followed by LightGBM (61.6%)
Ensemble methods significantly outperform deep learning models on this tabular dataset
LSTM shows the lowest performance (11.5% accuracy), indicating difficulty with frame-level classification
LightGBM shows the best stability (CV = 0.32%), followed by XGBoost (CV = 0.79%)

Figure 1. Overall Model Performance Comparison (7-Class Aggression)

Accuracy and Weighted F1 scores across 8 classification models with error bars (±1 SD)

Figure 2. Model Ranking by Accuracy

Sorted by mean accuracy (5 independent runs)

Figure 3. Model Stability Analysis

Coefficient of Variation (CV%) - Lower indicates more stable performance

Figure 4. Per-Class F1 Score Heatmap

F1 scores for each aggression class across all models (7 behavior classes)

Figure 5. Accuracy vs Weighted F1 Score Trade-off

Each point represents a model; bubble size indicates stability (smaller = more stable)

Table 1. Detailed Performance Metrics (7-Class Aggression Classification)

Mean ± standard deviation across 5 independent runs with different random splits