Analysis of Twitter Data on Healthcare Marketing by Text Mining and Machine Learning Techniques

Büşra Saylan; Songül Çınaroğlu

Makale detayları

Analysis of Twitter Data on Healthcare Marketing by Text Mining and Machine Learning Techniques

Yazarlar:

Büşra Saylan | Songül Çınaroğlu

Yayıncı:

Gazi University

Ana Başlıklar:

Yarı ve Denetimsiz Öğrenme | Makine Öğrenme (Diğer)

Dergi:

Bilişim Teknolojileri Dergisi Vol.17 Issue 2, pp.109-121

Belge Tipi:

Makale

Yayın Tarihi:

4/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001316477

pISSN:

1307-9697

Anahtar kelimeler:

Özet:

 TThe analysis of healthcare users moods regarding healthcare market plays a critical role in the analysis of perceptions towards the healthcare marketing. In this study, the views of people on healthcare marketing were evaluated using sentiment analysis with the data obtained from Twitter users. As a result of the sentiment analysis performed on 27079 English-language tweet data obtained from Twitter with API Key between October 1, 2022, and November 30, 2022, 50.36 n13638 of Twitter users opinions on healthcare marketing were positive, 32.21 n8721 were neutral and 17.43 n4720 were negative. According to the word clouds created, health, care, market, and insurance were determined as the most frequently repeated words. In order to find the best predictors of the mood of Twitter users, a new data set was created, and Neural Networks NN, Random Forest RF, Naive Bayes NB, and k-nearest neighbor k-NN were predicted by machine learning techniques. RF AUC0.707; CA0.646, and NN AUC0.706; CA0.645 were the best predictive machine learning techniques of Twitter users moods when compared to other machine learning techniques. With the Pythagorean tree, it was determined that the best predictor variable of Twitter users emotional states towards healthcare marketing was the word favorite. Future studies are suggested to be planned to cover a longer time to understand temporal changes better and provide a deeper perspective. Additionally, using a larger dataset is recommended to enhance the generalizability of analyses and contribute to the applicability of findings to a broader audience.

 TThe analysis of healthcare users moods regarding healthcare market plays a critical role in the analysis of perceptions towards the healthcare marketing. In this study, the views of people on healthcare marketing were evaluated using sentiment analysis with the data obtained from Twitter users. As a result of the sentiment analysis performed on 27079 English-language tweet data obtained from Twitter with API Key between October 1, 2022, and November 30, 2022, 50.36 n13638 of Twitter users opinions on healthcare marketing were positive, 32.21 n8721 were neutral and 17.43 n4720 were negative. According to the word clouds created, health, care, market, and insurance were determined as the most frequently repeated words. In order to find the best predictors of the mood of Twitter users, a new data set was created, and Neural Networks NN, Random Forest RF, Naive Bayes NB, and k-nearest neighbor k-NN were predicted by machine learning techniques. RF AUC0.707; CA0.646, and NN AUC0.706; CA0.645 were the best predictive machine learning techniques of Twitter users moods when compared to other machine learning techniques. With the Pythagorean tree, it was determined that the best predictor variable of Twitter users emotional states towards healthcare marketing was the word favorite. Future studies are suggested to be planned to cover a longer time to understand temporal changes better and provide a deeper perspective. Additionally, using a larger dataset is recommended to enhance the generalizability of analyses and contribute to the applicability of findings to a broader audience.

 TThe analysis of healthcare users moods regarding healthcare market plays a critical role in the analysis of perceptions towards the healthcare marketing. In this study, the views of people on healthcare marketing were evaluated using sentiment analysis with the data obtained from Twitter users. As a result of the sentiment analysis performed on 27079 English-language tweet data obtained from Twitter with API Key between October 1, 2022, and November 30, 2022, 50.36 n13638 of Twitter users opinions on healthcare marketing were positive, 32.21 n8721 were neutral and 17.43 n4720 were negative. According to the word clouds created, health, care, market, and insurance were determined as the most frequently repeated words. In order to find the best predictors of the mood of Twitter users, a new data set was created, and Neural Networks NN, Random Forest RF, Naive Bayes NB, and k-nearest neighbor k-NN were predicted by machine learning techniques. RF AUC0.707; CA0.646, and NN AUC0.706; CA0.645 were the best predictive machine learning techniques of Twitter users moods when compared to other machine learning techniques. With the Pythagorean tree, it was determined that the best predictor variable of Twitter users emotional states towards healthcare marketing was the word favorite. Future studies are suggested to be planned to cover a longer time to understand temporal changes better and provide a deeper perspective. Additionally, using a larger dataset is recommended to enhance the generalizability of analyses and contribute to the applicability of findings to a broader audience.

 Sağlık hizmeti kullanıcılarının sağlık hizmetleri pazarına ilişkin duygu durumlarının analizi sağlık hizmetleri pazarlamasına yönelik algıların analizinde kritik rol oynamaktadır. Bu çalışmada Twitter kullanıcılarından elde edilen verilerle kişilerin sağlık hizmetleri pazarlamasına yönelik görüşleri duygu analizi kullanılarak değerlendirilmiştir. Twitterdan API Key ile 1 Ekim 2022 ve 30 Kasım 2022 tarihleri arasında elde edilen 27079 İngilizce dilinde atılan tweet verileri üzerinde yapılan duygu analizi sonucunda Twitter kullanıcılarının sağlık hizmetleri pazarlaması konusundaki görüşlerinin 50,36sının n13638 olumlu, 32,21inin n8721 nötr ve 17,43ünün n4720 ise olumsuz olduğu belirlenmiştir. Oluşturulan kelime bulutlarına göre health, care, market ve insurance en sık tekrarlanan kelimeler olarak belirlenmiştir. Twitter kullanıcılarının duygu durumlarının en iyi tahmin edicilerini bulmak amacıyla yeni bir veri seti oluşturulmuş ve Yapay Sinir Ağları Neural Network-NN, Rastgele Orman Random Forest-RF, Naive Bayes NB ve k-en yakın komşu k-nearest neighbor k-NN makine öğrenmesi teknikleri ile tahmin edilmiştir. RF AUC0,707; CA0,646 ve NN AUC0,706; CA0,645 diğer makine öğrenmesi teknikleriyle karşılaştırıldığında Twitter kullanıcılarının duygu durumlarının en iyi tahmin edici makine öğrenmesi teknikleri olmuştur. Pisagor ağacı ile Twitter kullanıcılarının sağlık hizmetleri pazarlamasına yönelik duygu durumlarının en iyi tahmin edici değişkeninin favorite beğeni kelimesi olduğu ve kullanıcıların sağlık hizmetleri pazarlamasına yönelik atılan tweetlere çoğunlukla katıldığı belirlenmiştir. Gelecekteki yapılacak olan çalışmaların zaman içindeki değişiklikleri daha iyi anlamak ve derin bir perspektif sunmak amacıyla daha uzun bir zaman aralığını kapsayacak şekilde planlanması, ayrıca analizlerin genellenebilirliğini artırmak ve bulguların daha geniş bir kitleye uygulanabilir olmasına katkıda bulunmak için daha büyük bir veri seti kullanılması önerilmektedir.

Makale detayları

Analysis of Twitter Data on Healthcare Marketing by Text Mining and Machine Learning Techniques

Yazarlar:

Büşra Saylan | Songül Çınaroğlu

Yayıncı:

Gazi University

Ana Başlıklar:

Yarı ve Denetimsiz Öğrenme | Makine Öğrenme (Diğer)

Dergi:

Bilişim Teknolojileri Dergisi Vol.17 Issue 2, pp.109-121

Belge Tipi:

Makale

Yayın Tarihi:

4/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001316477

pISSN:

1307-9697

health care marketing

sentiment analysis

twitter

machine learning

pythagorean treesağlık hizmetleri pazarlaması

duygu analizi

twitter

makine öğrenmesi

pisagor ağacı

Sayı

Makale detayları

Yazarlar: Büşra Saylan | Songül Çınaroğlu

Yayıncı: Gazi University

Ana Başlıklar: Yarı ve Denetimsiz Öğrenme | Makine Öğrenme (Diğer)

Dergi: Bilişim Teknolojileri Dergisi Vol.17 Issue 2, pp.109-121

Belge Tipi: Makale

Yayın Tarihi: 4/2024

Belge Dili: Türkçe

Document ID (Org No.): 2001316477

pISSN: 1307-9697

Sayı

Yazarlar:

Büşra Saylan | Songül Çınaroğlu

Yayıncı:

Gazi University

Ana Başlıklar:

Yarı ve Denetimsiz Öğrenme | Makine Öğrenme (Diğer)

Dergi:

Bilişim Teknolojileri Dergisi Vol.17 Issue 2, pp.109-121

Belge Tipi:

Makale

Yayın Tarihi:

4/2024

Belge Dili:

Türkçe

Document ID (Org No.):

2001316477

pISSN:

1307-9697