Harnessing Gradient Boosting for Precision Date Classification: A Path to Sustainable Data-Driven Solutions
Ruchika Bhuria, Sheifali Gupta · 2024
This study presents a classification technique for date varieties using Gradient Boosting, a robust ensemble learning technique. The classification model is evaluated on a dataset comprising various date types, and the performance metrics are detailed through recall, precision and F1-scores for each class. The Gradient Boosting model exhibits high precision and recall across most classes, with notable performance for the DOKOL (F1-score: 0.95) and ROTANA (F1-score: 0.96) varieties, demonstrating its capability to distinguish these dates effectively. SAFAVI also indicates outstanding results with a high recall of 0.98, perfect precision of 1.00 . reflecting the model's performance in identifying this class. Conversely, the DEGLET and SOGAY classes present lower performance, with DEGLET achieving an F1-score of 0.68 and SOGAY 0.64, indicating areas where the model's predictive accuracy could be improved. The overall accuracy of the Gradient Boosting model is 89%, highlighting its strong classification capability across the dataset. The macro average F1-score of 0.85 underscores balanced performance, while the weighted average F1-score of 0.89 aligns closely with the overall accuracy, affirming the model's accuracy in distinguishing between different date varieties.