TensorFlow: GaussianNB - Fix Performance

[3C-SCSU]

🚨 Critical Performance Issue: GaussianNB Model Underperformance 📉 (60% Success Rate)

The Gaussian Naive Bayes (GaussianNB) classification model, recently merged via the referenced Pull Request, is exhibiting a concerning performance issue. Its current success rate stands at only 60%, which is significantly below expectations for our application.

---

📊 Performance Context

The poor performance of the GaussianNB model is highlighted when compared to existing classifiers that utilize the same high-quality data:

Classifier Model	Average Success Rate
Gaussian Naive Bayes (GNB)	60% (Suboptimal)
Random Forest	98%
Deep Learning Models	98%
Project Average (Target)	~98%

• Data Source: All models utilize the high-quality brainwave data provided by the Cloud Computing Club.

---

🔎 Call to Action: Investigation Required

Given the substantial performance difference, we need to immediately address and resolve the issues with the current GaussianNB implementation.

The community is requested to investigate the following potential causes:

1. Implementation Error: Is there a logical or technical error in the GNB algorithm's current codebase?
2. Data Suitability: Does the strict independence assumption of Naive Bayes fail catastrophically on our specific EEG features, limiting its theoretical maximum accuracy?
3. Optimization Needs: Does the model simply require intensive hyperparameter tuning, better feature scaling, or specific feature engineering/selection to perform comparably to the Random Forest and Deep Learning models?

Reference PR/Issue: #504