Neighbor Weighted K-Nearest Neighbor for Sambat Online Classification

Annisya Aprilia Prasanti (Universitas Brawijaya)
M. Ali Fauzi (Universitas Brawijaya)
Muhammad Tanzil Furqon (Universitas Brawijaya)

Indonesian Journal of Electrical Engineering and Computer Science
Vol 12, No 1: October 2018

Sambat Online is one of the implementation of E-Government for complaints management provided by Malang City Government.  All of the complaints will be classified into its intended department. In this study, automatic complaint classification system using Neighbor Weighted K-Nearest Neighbor (NW-KNN) is poposed because Sambat Online has imbalanced data. The system developed consists of three main stages including preprocessing, N-Gram feature extraction, and classification using NW-KNN. Based on the experiment results, it can be concluded that the NW-KNN algorithm is able to classify the imbalanced data well with the most optimal k-neighbor value is 3 and unigram as the best features by 77.85% precision, 74.18% recall, and 75.25% f-measure value. Compared to the conventional KNN, NW-KNN algorithm also proved to be better for imbalanced data problems with very slightly differences.