Klasifikasi Sentimen pada Twitter Terhadap WHO Terkait Covid-19 Menggunakan SVM, N-Gram, PSO

Noor Hafidz,Dewi Yanti Liliana

doi:10.29207/resti.v5i2.2960

Abstract

On March 2020 World Health Organization (WHO) has declared Covid-19 as global pandemic. As special agency of United Nation who responsible for international public healthy, WHO has done various actions to reduce this pandemic spreading rate. However, the handling of Covid-19 by WHO is not free from a number of controversies that gave rise to criticism and public opinion on the Twitter platform. In this research, a machine learning based classifier model has been made to determine the opinion or sentiment of the tweet. The dataset used is a set of tweets containing the phrase WHO and Covid-19 in period of March 1st until May 6th 2020 consisting of 4000 tweets with positive sentiments and 4000 tweets with negative sentiments. The proposed classifier model combined Support Vector Machine (SVM), N-Gram and Particle Swarm Optimization (PSO). The classifier model performance is evaluated using the value of Accuracy, Precision, Recall, and Area Under ROC Curve (AUC). Based on experiments conducted, the combination of SVM, N-gram (bigram), and PSO produced a pretty good performance in classifying tweet sentiment with values of Accuracy 0,755, Precision 0,719, Recall 0,837, and AUC 0,844.

Highlights

On March 2020 World Health Organization (WHO) has declared Covid-19 as global pandemic
Multiple Facebook posts claim that aspirin, lemon juice and honey have been combined to make a "home
Multiple Facebook posts claim that aspirin lemon juice and honey have been combined to make a home remedy for in Italy

Summary

Removing punctuation

Pada proses ini dilakukan penghapusan URL, user “aspirin_lemon” “lemon_juice” “juice_and”. Multiple Facebook posts claim that aspirin, lemon juice and honey have been combined to make a "home “make_a” “a_home” “for_in” “in_italy”}. Multiple Facebook posts claim that aspirin lemon juice kata-kata yang tidak memiliki dampak penting terhadap and honey have been combined to make a home remedy performa model classifier-nya [30]. Pada penelitian ini for in Italy digunakan Stopwords filtering bahasa Inggris dan multiple facebook posts claim that aspirin lemon juice and honey have been combined to make a home remedy jumlah huruf pada setiap kata yang dilewatkan berada di antara 4 dan 25 huruf. Evaluasi Model “misleading” “the” “has” “warned” “against” “for”} Metrik performa yang digunakan untuk melakukan. {“multiple” “facebook” “posts” “claim” “aspirin” evaluasi terhadap performa model classifier pada “lemon” “juice” “honey” “combined” “make” penelitian ini yaitu akurasi, presisi, recall, dan Area “home” “remedy” “claim” “misleading” “warned” under ROC curve (AUC). {“multiple” “facebook” “post” “claim” “aspirin” “lemon” “juice” “honey” “combine” “make” “home” “remedy” “claim” “mislead” “warn” “against”}

Ekstraksi Fitur

Pemodelan

Hasil dan Pembahasan

Kesimpulan analisis sentimen publik terhadap WHO tentang