COVID-19 pandemisinin ekonomi ve eğitim üzerindeki etkileri: Twitter üzerinden Türkiye örneği
Loading...
Date
2023
Authors
Çakmak, Elçin Timur
Journal Title
Journal ISSN
Volume Title
Publisher
Bursa Uludağ Üniversitesi
Abstract
COVID-19 pandemisi tüm dünyayı etkisi altına alan ve sosyal medyada büyük yankı uyandıran bir durum olmuştur. Bireyler COVID-19 hakkındaki düşüncelerini ve bu süreçte gerçekleşen olayları sosyal medya üzerinde takip etmeyi ve paylaşmayı tercih etmişlerdir. Son yıllarda kullanıcılar tarafından en popüler olarak değerlendirilen sosyal medya ortamının Twitter olduğu görülmektedir. Twitter üzerinden tweet atan kullanıcıların hangi duygularla fikirlerini beyan ettiklerini bulmak ilk bakışta pek mümkün değildir. Bu duyguların ortaya çıkarılması için Duygu Analizi yöntemi kullanılmaktadır. Literatürde sosyal medyada yer alan duyguların belirlenmesine yönelik çok sayıda çalışma bulunmaktadır. Ancak bu çalışmaların Türkiye’de ekonomi, turizm, eğitim, vb. alanlar ya da COVID-19 üzerine tek taraflı yapılmış duygu analizi çalışmaları olduğu görülmüştür. Türkiye’de COVID-19’un ekonomi ve eğitim için duygu analizi yöntemi kullanılarak analiz edildiği kapsamlı herhangi bir çalışmaya rastlanmamıştır. Çalışma kapsamında, dünyayı etkisi altına alan COVID-19 pandemisinin Türkiye’de ekonomi ve eğitim alanları üzerindeki etkisinin ulusal anlamda nasıl bir algı oluşturduğu ve pandemi dönemindeki ekonomi ve eğitim düzeylerinin pandemi öncesi döneme göre ne kadar değişiklik gösterdiğine dair sosyal medyada oluşturduğu etkinin ölçülmesi amaçlanmıştır. Bu amaçla, Türkiye’deki ekonomi ve eğitim süreçleri pandemi öncesi dönem ve pandemi dönemi olmak üzere başlıca 2 dönem altında incelenmiştir. Pandemi öncesi dönem olarak Mart 2019 – Şubat 2020 ve pandemi dönemi olarak ise Mart 2020 – Mayıs 2022 dönemi analize dahil edilmiştir. Pandemi döneminde yaşanan vaka sayılarının en çok gözlemlendiği “pik dönemleri” ve vaka sayılarındaki düşüşe bağlı olarak gerçekleşen “normalleşme dönemleri” ise pandemi döneminin alt dönemleri olarak analiz kapsamına alınmıştır. Söz konusu dönemlere ait Türkiye’de yaşayan kullanıcıların attığı Twitter üzerinden paylaştığı Türkçe tweetler veri setini oluşturmaktadır. Analiz kısmında söz konusu veriler; Twitter’dan Twitter API arayüzü ve Python programlama dili kullanılarak çekilmiştir. Elde edilen tweetler ham veri olarak değerlendirilmiş, ön işleme aşamasından geçirilerek temizlendikten sonra geriye kalan tweetler Duygu Analizi aşamasına dahil edilmiştir. Duygu Analizi yapabilmek için veriler içerisinde yer alan kelimeler olumlu, nötr ve olumsuz olarak gruplara ayrılarak duygu skorları elde edilmiştir. Duygu Analizi aşamasının ardından makine öğrenmesi ile analiz doğrultusunda yapılan sınıflandırmaların doğruluğunu ölçmek için sınıflandırma algoritmaları kullanılmıştır. Bunun için veri setinin %70’i eğitim ve %30’u test olacak şekilde iki gruba ayrılmıştır. Kullanılan makine öğrenmesi algoritmaları; Sınıflandırma ve Regresyon Ağaçları (CART), Naïve Bayes (NB), k-En Yakın Komşuluk (KNN) ve Rastgele Orman (RF) algoritmalarıdır. Elde edilen sonuçlara göre, model başarım ölçütleri olan kesinlik (P), doğruluk (A), duyarlılık (R) ve F-ölçütü (F) değerleri göz önüne alınarak kıyaslama yapılmıştır. Çalışmada Duygu Analizi sonuçlarına göre, Türkiye’deki ekonomik algı pandemi öncesi dönemde -0,2107 ve pandemi sonrası dönemde -0,22978 olarak elde edilmiştir. Buna göre; pandemi öncesi dönemde halihazırda ekonomi için olumsuz bir algı olduğu ortadadır. Pandemi sonrasında ise durumu çok da etkilemeyecek düzeyde olumsuzluğun biraz daha arttığı görülmektedir; ancak az da olsa artan olumsuz algının pandemiden kaynaklandığını düşünmek pek de mümkün gözükmemektedir. Bu durumun ekonominin mevcut durumdaki temel problemlerinden kaynaklandığı açıktır. Eğitim açısından değerlendirme yapıldığında ise, pandemi öncesi dönemde duygu analizi sonucu -0,2688 olarak elde edilmiştir. Türkiye’deki eğitim düzeyine olan algının da pandemi öncesi dönemde düşük düzeyde olumsuz olduğu söylenebilir. Pandemi döneminde bu duygunun daha da negatifleşerek -0,5091 seviyesine çıktığı görülmüştür. Bu durum pandemi döneminde insanların eğitime karşı olan beklentilerinin karşılanmadığını, eğitim kalitesinin pandemi öncesi döneme göre daha da düştüğünü, bu nedenle olumsuz algının bu dönemde daha da arttığını ortaya koymaktadır.
The COVID-19 pandemic has been a situation that has affected the whole world and has had great repercussions on social media. Individuals preferred to follow and share their thoughts about COVID-19 and the events that took place in this process on social media. It is seen that Twitter is the most popular social media medium by users in recent years. At first glance, it is not possible to find out with what feelings the users tweeting on Twitter express their opinions. Sentiment Analysis method is used to reveal these emotions. There are many studies in the literature on the determination of emotions in social media. However, these studies in Turkey economy, tourism, education, etc. It has been observed that there are unilateral sentiment analysis studies on areas or COVID-19. No comprehensive study was found in Turkey in which COVID-19 was analyzed using sentiment analysis for economy and education. Within the scope of the study, it is aimed to measure the impact of the COVID-19 pandemic, which has affected the world, on the economy and education fields in Turkey, how it creates a perception in the national sense and how much the economy and education levels in the pandemic period have changed compared to the pre-pandemic period. For this purpose, the economy and education processes in Turkey were examined under two main periods: the pre-pandemic period and the pandemic period. The period between March 2019 – February 2020 as the pre-pandemic period and March 2020 – May 2022 as the pandemic period were included in the analysis. “Peak periods” in which the number of cases experienced during the pandemic period are observed the most, and “normalization periods” due to the decrease in the number of cases are included in the analysis as sub-periods of the pandemic period. Turkish tweets shared on Twitter by users living in Turkey belonging to the mentioned periods constitute the data set. In the analysis part, the data in question was captured using the Twitter API interface and the Python programming language. The tweets obtained were evaluated as raw data, and the remaining tweets were included in the Sentiment Analysis stage after they were cleaned through the preprocessing stage. In order to perform Sentiment Analysis, the words in the data were divided into groups as positive, neutral and negative, and emotion scores were obtained. After the Sentiment Analysis stage, classification algorithms were used to measure the accuracy of the classifications made in line with the analysis with machine learning. For this, the data set was divided into two groups, 70% training and 30% testing. Machine learning algorithms used; Classification and Regression Trees (CART), Naïve Bayes (NB), k-Nearest Neighborhood (KNN) and Random Forest (RF) algorithms. According to the results obtained, a comparison was made considering the model performance criteria of precision (P), accuracy (A), sensitivity (R) and F-criterion (F). According to the results of Sentiment Analysis in the study, the economic perception in Turkey was -0.2107 in the pre-pandemic period and -0.22978 in the post-pandemic period. According to this; It is clear that there is already a negative perception for the economy in the pre-pandemic period. After the pandemic, it is seen that the negativity has increased a little more at a level that will not affect the situation much; however, it does not seem possible to think that the slightly increased negative perception is due to the pandemic. It is clear that this situation is due to the basic problems of the economy in the current situation. When evaluated in terms of education, the emotion analysis result in the pre-pandemic period was -0.2688. It can be said that the perception of the level of education in Turkey was negative at a low level in the pre-pandemic period. During the pandemic period, it was observed that this feeling became even more negative and increased to the level of -0.5091. This situation reveals that people's expectations towards education were not met during the pandemic period, the quality of education decreased even more compared to the pre-pandemic period, and therefore the negative perception increased even more in this period.
The COVID-19 pandemic has been a situation that has affected the whole world and has had great repercussions on social media. Individuals preferred to follow and share their thoughts about COVID-19 and the events that took place in this process on social media. It is seen that Twitter is the most popular social media medium by users in recent years. At first glance, it is not possible to find out with what feelings the users tweeting on Twitter express their opinions. Sentiment Analysis method is used to reveal these emotions. There are many studies in the literature on the determination of emotions in social media. However, these studies in Turkey economy, tourism, education, etc. It has been observed that there are unilateral sentiment analysis studies on areas or COVID-19. No comprehensive study was found in Turkey in which COVID-19 was analyzed using sentiment analysis for economy and education. Within the scope of the study, it is aimed to measure the impact of the COVID-19 pandemic, which has affected the world, on the economy and education fields in Turkey, how it creates a perception in the national sense and how much the economy and education levels in the pandemic period have changed compared to the pre-pandemic period. For this purpose, the economy and education processes in Turkey were examined under two main periods: the pre-pandemic period and the pandemic period. The period between March 2019 – February 2020 as the pre-pandemic period and March 2020 – May 2022 as the pandemic period were included in the analysis. “Peak periods” in which the number of cases experienced during the pandemic period are observed the most, and “normalization periods” due to the decrease in the number of cases are included in the analysis as sub-periods of the pandemic period. Turkish tweets shared on Twitter by users living in Turkey belonging to the mentioned periods constitute the data set. In the analysis part, the data in question was captured using the Twitter API interface and the Python programming language. The tweets obtained were evaluated as raw data, and the remaining tweets were included in the Sentiment Analysis stage after they were cleaned through the preprocessing stage. In order to perform Sentiment Analysis, the words in the data were divided into groups as positive, neutral and negative, and emotion scores were obtained. After the Sentiment Analysis stage, classification algorithms were used to measure the accuracy of the classifications made in line with the analysis with machine learning. For this, the data set was divided into two groups, 70% training and 30% testing. Machine learning algorithms used; Classification and Regression Trees (CART), Naïve Bayes (NB), k-Nearest Neighborhood (KNN) and Random Forest (RF) algorithms. According to the results obtained, a comparison was made considering the model performance criteria of precision (P), accuracy (A), sensitivity (R) and F-criterion (F). According to the results of Sentiment Analysis in the study, the economic perception in Turkey was -0.2107 in the pre-pandemic period and -0.22978 in the post-pandemic period. According to this; It is clear that there is already a negative perception for the economy in the pre-pandemic period. After the pandemic, it is seen that the negativity has increased a little more at a level that will not affect the situation much; however, it does not seem possible to think that the slightly increased negative perception is due to the pandemic. It is clear that this situation is due to the basic problems of the economy in the current situation. When evaluated in terms of education, the emotion analysis result in the pre-pandemic period was -0.2688. It can be said that the perception of the level of education in Turkey was negative at a low level in the pre-pandemic period. During the pandemic period, it was observed that this feeling became even more negative and increased to the level of -0.5091. This situation reveals that people's expectations towards education were not met during the pandemic period, the quality of education decreased even more compared to the pre-pandemic period, and therefore the negative perception increased even more in this period.
Description
Keywords
Duygu analizi, Makine öğrenmesi, Covid-19, Pandemi, Ekonomi, Eğitim, Türkiye, Twitter, Sentiment analysis, Machine learning, Pandemic, Economics, Education, Turkey
Citation
Çakmak, E. T. (2023). COVID-19 pandemisinin ekonomi ve eğitim üzerindeki etkileri: Twitter üzerinden Türkiye örneği. Yayınlanmamış doktora tezi. Bursa Uludağ Üniversitesi Sosyal Bilimler Enstitüsü.