Sentiment Analysis of Delivery Service Opinions on Twitter Documents using K-Nearest Neighbor

Abstract

The global pandemic of Covid-19 in 2020 encourage a significant increase of online trading such as selling and buying around the world, especially in Indonesia. The dependency on goods delivery services to fulfill transaction process among sellers and buyers are essential as the government policy enforce a limitation of errands outside home. Since 2000, top brand awards have often conducted official surveys in terms of awarding and displaying comparisons of goods or services based on public opinion from several regions. One of the services that became the top brand survey is delivery service. However, it was found that the survey rankings were not quite accurate because the public and companies were still unable to find out in detail the success factors and drawbacks of their brands based on the ratings submitted. This research was conducted to analyze opinion sentiment of delivery service (JNE, J&T, TIKI, Pos Indonesia, and DHL Indonesia) and determine the aspect of timeliness and quality. The data is retrieved from Twitter “tweets” of delivery service official account of @JNE_ID, @shippexpressid, @PosIndonesia, and @IdTiki, and search keyword of “DHL Indonesia”. K-Nearest Neighbor with parameter K=3 is adopted as the main algorithm following some text processing method, feature extraction and feature selection. The result shows DHL Indonesia is in the first position at 55.6% of Twitter users who gave positive feedback for service aspect by timeliness and at 15% for service aspect by quality. This research system has accuracy at 94.56%, precision at 62.31%, recall at 45.26%, and F-measure at 62.32%. It can be concluded that the parameters used are quite optimal.