IMPLEMENTASI SOFT VOTING CLASSIFIER UNTUK PREDIKSI KINERJA MITRA (STUDI KASUS : PT TELKOM INDONESIA) 

The performance of partners in the Regional Division VII Makassar at PT Telkom Indonesia is evaluated based on their contributions to the project work of procuring and installing the Outside Plant Fiber Optic (OSP-FO), which involves installing fiber optic networks outside buildings in the telecommunications area of Makassar and extending to the eastern parts of Indonesia. This evaluation encompasses various aspects, including General Affairs (GA), Planning Engineering Development (PED), Supply Management Information for Logistic Enhancement (SMILE), and field assessments by the Telecommunications Area (Witel). To generate performance predictions, this study implements the soft voting classifier method. In this research, the purpose of implementing the soft voting classifier method, which combines the XGBoost, AdaBoost, and gradient boosting algorithms, is to improve the accuracy of partner performance predictions. The results of this method are used in creating an information system for the evaluation process of all partner performances. In the soft voting classifier model, the firefly method is applied to determine the optimal weights for the three algorithms used. These optimal weights are applied to combine the classification results from each algorithm in making the final prediction. This study uses data from 7329 projects from 39 partners. The results show that partner performance predictions with the soft voting classifier method have better performance, achieving an accuracy of 91.6%, compared to the accuracy of the XGBoost algorithm at 87.1%, the AdaBoost algorithm at 81.2%, and the gradient boosting algorithm at 86.7%.