Boosting Methods Comparison : XGBoost, Adaboost and Gradient Boosting for Business Partner Performance Prediction
XGBoost is known for its superiority in handling large datasets and providing regularization that reduces overfitting, AdaBoost improves model accuracy by adjusting weights on misclassified instances in previous iterations, and Gradient Boosting builds models sequentially to minimize the loss function.
Abstract
Measuring the performance of business partners is a crucial technique for overseeing and sustaining an organization or company's competitive edge. This process involves contrasting anticipated results with actual results, examining deviations from the plan, assessing individual performance, and evaluating the progress toward meeting set goals. Additionally, partnership is described as a collaboration between two or more parties based on a mutual agreement to benefit each other and increase revenue. In this study, each algorithm was tested to evaluate its ability to classify business partners based on their performance in completed projects. The research process includes several key stages, starting with data collection and preprocessing, which involves feature selection, data cleaning, data transformation, and data labeling. Following that, the study focuses on the application of algorithms, where XGBoost is known for its superiority in handling large datasets and providing regularization that reduces overfitting, AdaBoost improves model accuracy by adjusting weights on misclassified instances in previous iterations, and Gradient Boosting builds models sequentially to minimize the loss function. Evaluation of the XGBoost model showed the best performance in partner classification compared to other models, with Precision value 86.8%, Recall 76%, F1-Score 86.9%, and Accuracy 87.1%.