TR
Avrasya Ekonometri �statistik ve Ampirik Ekonomi DergisiYl:2020 Say: 17 Alan: alma ktisad ve ktisadi Demografi

Engin KARAMAN, idem ARICIGL LAN
COVD-19 DNEMNDE TURZM LE LGL TWTLERN MAKNE RENMES YNTEMLERYLE DUYGU ANALZ
 
Tm dnyada etkili olan ve kresel bir pandemi zellii tayan COVD-19 virs, Trkiyede ki turizm tercihleri zerinde de etkisini gstermitir. Bu almada, Nisan- Austos 2020 tarihleri arasnda atlan turizm hahstagli (#turizm) Trke twitler zerinden duygu analizi almas yaplmtr. Veriler Twitter API uygulamasndan elde edilmitir. Bu srete toplanan 9678 adet ileti gerekli n ileme ve dntrme sreleri zerinden yaplandrlarak 4202 adet olarak analize hazr hale getirilmi ve iletiler anlamsal adan ierdikleri duygu ifadesine gre kategoride etiketlenmitir. Duygu analizi almalarnda Makine renmesinin en ok kullanlan yntemlerinden (algoritmalarndan); Lojistik Regresyon Analizi, Karar Aac, Multinominal Naive Bayes Analizi, Kmeleme Analizi(k-En Yakn Komu), Destek Vektr Makineleri ve Rassal Ormanlar kullanlarak snflandrma performanslar karlatrlm ve en baarl model olarak 0.66 doruluk skoruyla Lojistik Regresyon modeli olmutur. Oluturulan model gelitirilmeye ak olmakla birlikte tahminlime almalarnda kullanma uygundur.

Anahtar Kelimeler: Turizm, Python, Twitter API, Duygu Analizi, Makine renmesi


SENTIMENT ANALYSIS OF TOURISM-RELATED TWEETS DURING COVID-19 OUTBREAK THROUGH MACHINE LEARNING TECHNIQUES
 
Covid-19 virus which is effective all the world and is a global pandemic also affected tourism choices in Turkey. In this study, sentiment analysis study was conducted over the tourism hahstagli (#turizm) Turkish tweets posted between April and August 2020. The data was obtained from the Twitter API application. 9678 messages collected in this process were structured over the necessary pre-processing and transformation processes and made ready for analysis as 4202 messages, and the messages were labeled in three categories (neutral, positive and negative) according to the emotion expressions they contain.Classification performances were compared using Machine Learning algorithms (Logistic Regression Analysis, Decision Tree, Multinominal Naive Bayes Analysis, Cluster Analysis (k-Nearest Neighbor), Support Vector Machines and Random Forests), which are frequently used in sentiment analysis studies. As a result, Logistic Regression model was found to be the most successful model.

Keywords: Tourism, Python, Twitter API, Sentiment Analysis, Machine Learning


Detay

ÇERK