Analisa Sentimen Tweet Berbahasa Indonesia dengan Menggunakan Metode Pembobotan Hybrid TF-IDF pada Topik Transportasi Online

Abstract

Abstrak Beberapa tahun belakangan ini, muncul perusahaan-perusahaan penyedia jasa transportasi yang menggunakan aplikasi berbasis android dalam proses pelayanannya atau biasa disebut dengan transportasi online. Hal ini dilakukan untuk meningkatkan pelayanan terhadap pengguna jasa transportasi. Hadirnya transportasi online seperti Gojek, Grab dan Uber menimbulkan masalah sosial antara supir transportasi online dan supir transportasi non aplikasi. Penyebabnya dikarenakan sebagian besar masyarakat beralih menggunakan transportasi online, sehingga pendapatan supir transportasi non aplikasi menurun. Pada penelitian ini, dilakukan analisa sentimen terhadap tweet berbahasa Indonesia tentang transportasi online dengan menggunakan metode pembobotan Hybrid TF-IDF dan kNN sebagai metode klasifikasinya. Hasil terbaik dari pengujian cross validation pada uji variable k adalah k=5 dengan nilai akurasi 70%, presisi kelas positif 68%, presisi kelas negatif 75%, recall kelas positif 82%, recall kelas negatif 59%, f-measur kelas positif 74% dan f-measure kelas negatif 65%.Abstract In recent years, there are same new companies which uses android applications in their service process or commonly called with online based transportation. This case do for improve their service to online based transportations customers. The presence of online based transportation like Gojek, Grab, and Uber inflict social problem between online based transportations drivers with non-aplication based transportations drivers. The is due to most of the people change over to online based transportation, so non-aplication based transportations drivers income decreased. In this research, sentiment analysis against Indonesian tweets about online based transportation by using Hybrid TF-IDF weighting method and kNN classification method. The best cross validation result on the test k variable are k=5 with accuracy value 70%, positive class precision 68%, negative class precision 75%, positive class recall 82%, negative class recall 59%, positive class f-measure 74%, and negatif class f-measure 65%.

Classification of Traffic Congestion in Indonesia Using the Naive Bayes Classification Method

“…In this weighting, each Twitter dataset that has been preprocessed will be given a weighting value using TF-IDF which will produce a value for each text dataset from the tweet (Wahyunita et al, 2020).…”

Comparison of Classification Algorithms for Predicting Indonesian Fake News using Balanced and Imbalanced Datasets

“…In the dataset obtained, not all letters are consistent with lower case or lower case. This Case Folding stage is needed in converting the entire text in the dataset into a standard form or into lowercase letters by using the lower() function to generalize text that is not yet in lowercase form [14]. The results of the case folding process are presented in Figure 3.…”

“…The tokenization stage is used to separate a sentence into word for word which is usually called a token so that it can be analyzed and facilitate data processing in the next stage [14]. The results of the tokenizing process are presented in Figure 5.…”

“…The stemming stage is done to change the word into its basic word form by removing the initial and final suffixes. The stemming process uses a special library for Indonesian language processing, namely the Python Sastrawi library [14]. The results of the stemming process are presented in Figure 7.…”

Implementasi Bidirectional LSTM untuk Analisis Sentimen Terhadap Layanan Grab Indonesia

“…Hasil penelitian menunjukkan polarisasi sentimen positif 44,51% dan negatif 45,80%. Dalam penelitian yang dilakukan Wahyunita [7] tentang transportasi online, analisis sentimen dilakukan pada tweet berbahasa Indonesia dengan metode pembobotan Hybrid TF-IDF dan metode algoritma K-Nearest Neighbor (KNN) sebagai metode klasifikasi sentimen. Dengan menggunakan pengujian cross validation mendapatkan hasil terbaik dengan nilai akurasi 70% untuk k=5, presisi kelas sentimen positif 68%, negatif 75%, recall kelas sentimen positif 82%, negatif 59%, F-measure kelas sentimen positif 74% dan negatif 65%.…”