İş zekası ve duygu analizi: Sakarya merkezli sosyal medya verilerinin doğal dil işleme yaklaşımlarıyla incelenmesi = Business intelligence and sentiment analysis: Examining Sakarya-centric social media data through natural language processing approaches

Abstract

Bu çalışma, sosyal medya kullanımının gün geçtikçe arttığı ve yaygınlaştığı bir döneme odaklanmaktadır. Özellikle Twitter gibi dinamik bir platformdan elde edilen sürekli güncellenen ve hızla artan metinsel veriler üzerinde yoğunlaşarak, bu verilerin analizini gerçekleştirmeyi amaçlamaktadır. Bu metin verileri, doğal dil işleme yöntemleri ve makine öğrenmesi algoritmaları kullan ılarak madencilik yapılarak incelenmekte ve bu sayede sosyal medya ortamını, toplumu ve çevreyi oluşturan bireylerin düşünce, his ve duygularını anlaşılır bir şekilde ortaya koymayı hedeflemektedir. Temel olarak bu çalışma, metinsel verilerin derinlemesine incelenmesi ve analiz edilmesi yoluyla, sosyal medya kullanıcılarının zihinsel dünyalarını daha anlamlı bir şekilde ortaya çıkarmayı amaçlamaktadır. Twitter gibi duygu ve düşüncelerin sıkça paylaşıldığı bir platform üzerinde gerçekleştirilen bu çalışmada, Tweet metinleri üzerinde gerçekleştirilen duygu analizi ve kategori tahminleri sayesinde elde edilen veriler, iş zekası mimarisi kullanılarak anlamlandırılmış ve çeşitli grafiklerle görselleştirilmiştir. Bu yaklaşım, geleneksel anket çalışmalarının yerine geçerek daha düşük maliyetle, hızlı erişim ve sonuçlar sunabilmeyi amaçlamaktadır. Twitter üzerinden elde edilen bu verilerin analizi, duygu ve düşüncelerin toplumda nasıl yayıldığını anlamak ve daha etkili bir veri toplama yöntemi sunmak amacıyla gerçekleştirilmiştir. Bu araştırma, çeşitli yazılımlar ve yaklaşımların bir araya getirildiği bir çerçeve içinde gerçekleşmiş olup, uzun vadeli çıkarımlara ulaşmayı hedeflemektedir. İş zekası mimarisi ve araçlarıyla otomatikleştirilen çalışmanın adımları, belirli sonuçlara ve başarımlara ulaşmak için özenle düzenlenmiştir. Performans, optimum düzeyde tutularak veri kaybı ve yanlışlık oranı en düşük seviyeye çekilmiştir. Veri koruma ve yedekleme süreçleri için güvenilir veri tabanı sistemleri tercih edilmiştir. Bu şekilde, araştırmanın güvenilirliği artırılmış ve elde edilen verilerin güvenliği sağlanmıştır. Temizlenmiş ve bilgiye dönüştürülmüş veriler, çeşitli veri görselleştirme uygulamaları ve çözümleri kullanılarak anlamlı bir görsel forma getirilmiş ve bu önemli bilgiler, etkili görsel raporlar şeklinde sunulmaktadır. Bu raporlar, karmaşık veri setlerini anlamak ve anahtar bulguları hızlıca görmek isteyenlere verileri anlaşılık kılmak amacıyla tasarlanmıştır. Araştırmanın temel amacı, sadece geçici sonuçlara değil, aynı zamanda sürdürülebilir bir şekilde uzun vadeli çıkarımlara ulaşmaktır. Bu nedenle, verilerin anlamlandırılması ve görselleştirilmesi, araştırmanın gelecekteki analizlere ve karar süreçlerine değer katmasını sağlamak üzere özenle planlanmıştır. Bu görsel sunumlar, araştırma sürecinin etkili bir iletişim aracı olup, bulguların geniş bir izleyici kitlesine kolayca aktarılmasını amaçlamaktadır.

This study focuses on a period in which the use of social media is increasing and becoming more widespread. Specifically, it aims to analyze the continuously updated and rapidly increasing textual data obtained from dynamic platforms like Twitter. These text data are examined through natural language processing methods and machine learning algorithms, aiming to understand the thoughts, feelings, and emotions of individuals shaping the social media environment, society, and the community. Fundamentally, the study aims to uncover the mental worlds of social media users more meaningfully by in-depth examination and analysis of textual data. In this context, the application of techniques like sentiment analysis and social media analytics functions as a powerful toolkit that allows for both a deeper understanding and future prediction of user interactions and behaviors on different social media platforms. The ultimate goal of this investigation is to make sure that the knowledge extracted from social media data through the use of these advanced methods reaches a degree of depth and coverage necessary for well-informed decision-making and academic research. Focused mostly on social media platforms where emotions and thoughts are frequently expressed, such as Twitter, this study involved careful data pretreatment efforts to enable sentiment analysis and category predictions in tweet messages. During this preparation stage, a methodical technique was used to thoroughly clean the text data. This included filtering out irrelevant phrases, eliminating uncertainties, and removing superfluous letters. The cleaned and refined dataset was then carefully selected to enhance the applicability and effectiveness of machine learning algorithms in the later stages of data analysis and prediction. Using techniques like the Turkish BERT algorithm, which is well-known for having better performance metrics than other algorithms, the cleaned and documented dataset was processed once again. Subsequent data transformation and modeling processes were included into the business intelligence architecture framework in an elegant manner, resulting in data that was carefully prepared for visualization. These deliberate actions serve to improve the effectiveness of business intelligence-driven analysis in addition to enabling a more thorough understanding of insights obtained from Twitter data. The data moved smoothly to the Power BI application, where it was carefully transformed into a multitude of visual representations through the establishment of interrelationships. These images were carefully selected and arranged on several pages labeled "Location Analysis," "Category Analysis," and "Sentiment Analysis." Notably, a detailed analysis was carried out under these specific parts, concentrating on the connections and discussion regarding shares that come from the province of Sakarya. In the sentiment analysis section, special attention was paid to identifying the dominant feelings expressed in shares written by users from Sakarya province as well as a detailed examination of how they were distributed. In the meantime, the category analysis went into explaining the themes contained in the shares and also looked closely at how they were distributed among categories. The location analysis section further clarified any apparent correlations with particular theme motifs and illuminated the spatial distribution of shares within Sakarya province. Essentially, the visual analyses selected by Power BI provide a comprehensive understanding of the sharing preferences and behavioral patterns displayed by Sakarya province's social media users. The category analysis results explore the shared content's thematic landscape in Sakarya province and indicate a significant focus on topics related to "Politics" and "Culture." In these theme domains, a tendency toward "Very Positive" and "Neutral" attitudes was observed, highlighting Sakarya users' tendency to adopt a positive or neutral manner, especially when participating in conversations about political and cultural issues. These observations provide light on the general mindset of Sakarya province's social media users, suggesting a tendency to take a favorable or impartial viewpoint while discussing issues of political and cultural importance. After analyzing sentiment analysis in more detail, several different emotions, such as "Very Positive," "Neutral," and "Negative," were shown to be common in shared content. But when one looked closely at measures like retweet and like counts, one interesting finding emerged: shares that elicited happy feelings received more interaction. This observation implies that users are more sensitive to content with a positive tone, as seen by their active desire to promote and spread such content on social media in Sakarya province. As a result, the results highlight the tendency of Sakarya province social media users to favor content that receives good feedback, highlighting a tendency to actively interact with and spread positive information. The geography study produced some interesting findings: within Sakarya province, the districts with the highest number of tweet shares were Adapazarı and Serdivan. Even though the number of tweets from each of these districts varied, it was clear that shares from Serdivan received more attention and retweets. This phenomena points to the existence of a social media user base in Serdivan that is more vibrant and interaction-driven. These results highlight the potential impact of regional dynamics—which go beyond simple physical proximity—on the degree of participation among social media users. With the ultimate goal of accelerating the acquisition of findings, this technique aims to replace traditional survey methodologies with more economical and efficient alternatives. The study's stated goal is to use social media data to better understand the mechanisms behind how ideas and emotions spread throughout society and to provide a more effective means of gathering data. Through the use of social media platforms to enable real-time data gathering and analysis, this method makes it possible to quickly and thoroughly generate insights, which in turn makes it possible to more quickly and thoroughly understand the emotional and intellectual dynamics of society. Here, the goal is to avoid the time and cost constraints that come with conventional survey approaches while attempting to extract meaningful stories from large datasets. Examining Twitter data provides a broad overview of general trends, emotional responses, and mental processes, providing a deeper and more complex understanding. At the heart of this methodology is the effort to examine social media users' real-time reactions, which promotes agility and quickness in the data collection process and makes the information-gathering process more dynamic and effective. Furthermore, by addressing the always changing and expanding data pool on social media sites, this method enables more intelligent monitoring and understanding of the changing emotional and cognitive fabric of society. As a result, these data collecting and analysis procedures take on a more flexible and modern color, providing decision-makers with more accurate and fast information. In order to obtain long-lasting insights, this study has been conducted inside a comprehensive framework that integrates multiple software tools and approaches. The business intelligence architecture and technologies were used to carefully plan and carry out the automated process's methodical steps. Efforts were undertaken throughout this continuum to reduce error rates and data loss, which in turn guaranteed the accuracy and dependability of the analyzed data and strengthened the credibility of the generated results. In addition, the integration of integrated software tools and procedures has expanded the scope of the research, providing a deeper knowledge by careful examination. This enhancement improves the ability to accomplish long-term goals and make more effective decisions. The need for dependable database systems in data backup and protection procedures emphasizes how crucial it is to guarantee the security of the collected data. By maintaining the highest standards throughout the data analysis process, this robust technique strives to facilitate the achievement of long-term perspectives and encourage confidence in the obtained insights. This study presents cleaned and converted data in an insightful visual format using visually striking reports that are easily incorporated into the business intelligence architecture and natural language processing techniques. The value of these data visualizations is enhanced by utilizing business intelligence architecture, which expedites the analysis process and allows the gained insights to be evaluated from a broader perspective. As a result, the research findings are presented in a way that makes them easier to understand, which increases their effectiveness in guiding decision-making processes.