Customer Classification Using Naive Bayes Classifier With Genetic Algorithm Feature Selection

Abstract

There is a tendency to decrease the number of speedy customers in the operational area of ​​North Sumatra due to customer dissatisfaction. Termination of employment is carried out by the customer against PT. Telekomunikasi Indonesia, Tbk in North Sumatra. There is no management of customer data classification so that classification information based on certain product purchases cannot be known. Naïve Bayes is a classification algorithm that is easy to use but has weaknesses which result in poor performance, therefore feature selection is needed, the genetic algorithm is an algorithm that is able to select attributes in research, will be selected based on the highest weight so that the accuracy of the prediction results is more optimal. The steps taken in the measurement model using the Naive Bayes Classifier (NBC) approach and the model using the GA-NBC approach obtained accurate results from cross validation measurements, Confusion Matrix, ROC curves for the classification of existing and speedy telephone subscribers. The stages of the Naive Bayes process are: data collection, data preprocessing, processing of the Naive Bayes Classifier algorithm. Then the results are validated and evaluated using the Text Mining Algorithm, and calculating the parameters based on the genetic algorithm. The accuracy produced by the Naive Bayes Classifier model is 85.08%. The accuracy produced by the Naive Bayes Classifier model with the selection of Genetic Algorithm features increased to 89.31%.

Prediction of Stunting in Toddlers Combining the Naive Bayes Method and the C4.5 Algorithm

“…In research conducted (Nasution, Dar, & Nasution, 2023) the naive Bayes method was used to classify students' interest in using gaming laptops and the accuracy obtained was 90%, this means that the naive Bayes method can be used to carry out classification well (F. F. Hasibuan, Dar, & Yanris, 2023). This method is very effective and efficient for large datasets, making it popular in the fields of natural language processing and pattern recognition (Supendar, Rusdiansyah, Suharyanti, & Tuslaela, 2023) (Siregar, Irmayani, & Sari, 2023) (Tanjung, Tampubolon, Panggabean, & Nandrawan, 2023). Naive Bayes is easy to implement and requires a smaller amount of training data to estimate parameters, so it is effective for applications that require fast responses (Madjid, Ratnawati, & Rahayudi, 2023) (Anam, Rahmiati, Paradila, Mardainis, & Machdalena, 2023).…”