Naive Bayes classifier optimization for text classification on e-government using particle swarm optimization

One of the public e-government services is a web-based online complaints portal. Text of complaint needs to be classified so that it can be forwarded to the responsible office quickly and accurately. The standard classification approach commonly used is the Naive Bayes Classifier (NBC) and k-Nearest Neighbor (k-NN), which still classifies one label and needs to be optimized. This research aims to classify the complaint text of more than one label at the same time with NBC, which is optimized using Particle Swarm Optimization (PSO). The data source comes from the Sambat Online portal and is divided into 70 % as training data and 30 % as testing data to be classified into seven labels. NBC and k-NN algorithms are used as a comparison method to find out the performance of PSO optimization. The 10-fold cross-validation shows that NBC optimization using PSO achieves an accuracy of 87.44 % better than k-NN of 75 % and NBC of 64.38 %. The optimization model can be used to increase the effectiveness of services to e-government in society.