Puji Dwi Rinanda, -
(2024)
IMPLEMENTATION OF PNN, ANN, AND K-NN ALGORITHMS ON INDONESIAN MARKETPLACE REVIEWS ON GOOGLE PLAY STORE.
International Conference in Emerging Technologies for Sustainability and Intelligent Systems (ICETSIS).
Abstract
The Google Play Store is an official application owned by Google that provides digital content such as games, applications, movies, music, and books with various categories. One of the applications available on the Play Store is a buying and selling or marketplace service application, namely Shopee, Tokopedia, Lazada, Blibli, and Bukalapak. These marketplace applications have many reviews or responses from users regarding the issues they experience while shopping, such as fraud, delayed delivery, mismatched items received, expensive shipping costs, damaged items received by customers, or other problems. These reviews can be analyzed using text mining techniques. In this research, three classification techniques, namely Artificial Neural Network (ANN), Probabilistic Neural Network (PNN), and K-Nearest Neighbor (KNN). The training and test data are divided using the 10-fold Cross Validation technique, with 1000 data points used for each marketplace. The review classes considered are positive, negative, and neutral. In the experiments with the data, the highest accuracy is achieved by the ANN algorithm. ANN outperforms PNN and K-NN, with an accuracy of 81.54%, while PNN has an accuracy of 69.32%, and K-NN has an accuracy of 68.4%. Despite having the highest accuracy, ANN requires a relatively long training time compared to PNN and K-NN. However, in terms of performance, ANN is better than PNN and K-NN in modeling the five datasets.
Actions (login required)
 |
View Item |
