آرشیو

آرشیو شماره‌ها:
۱۲۷

چکیده

هدف این پژوهش شناسایی و دسته بندی نظرات کاربران فارسی زبانِ «توئیتر» در رابطه با موضوع بحران های خشکسالی در ایران و سپس، توسعه مدلی برای تشخیص این نظرات در پلتفرم «توئیتر» است. ازاین رو، مدلی با کمک روش های یادگیری ماشین و متن کاوی برای تشخیص نظرات کاربران فارسی زبانِ «توئیتر» توسعه داده شد. جامعه آماری پژوهش، تعداد 42028 توئیت منتشرشده در بازه زمانی یک ساله مرتبط با خشکسالی بود. این توئیت ها با کلیدواژه های مرتبط با مسایل و بحران های خشکسالی در ایران، از «توئیتر» استخراج شد و سپس، یک نمونه 2300تایی توئیت به روش کیفیِ تحلیل تم، برچسب گذاری، دسته بندی و تحلیل شد. آنگاه یک دسته بندی چهارتایی از نظرات کاربران در رابطه با بحران های خشکسالی و تاب آوری ایرانیان در برابر این بحران ها شناسایی گردید. سپس، مبتنی بر این چهار دسته، مدل یادگیری ماشین بر اساس روش رگرسیون لجستیک برای پیش بینی و تشخیص انواع نظرات در پست های «توئیتر»، آموزش داده شد. مدل توسعه داده شده دارای دقت 09/66 درصد و معیار افِ 60 درصد است که نشان می دهد این مدل، برای تشخیص نظرات کاربران ایرانی در ارتباط با بحران های خشکسالی از عملکرد خوبی برخوردار است. تشخیص نظرات در رابطه با بحران های خشکسالی در پلتفرم های اجتماعی مانند «توئیتر» از نیازهای سیاست گذاران و مدیران این حوزه است. توسعه مدل تشخیص این نظرات با روش های یادگیری ماشینی می تواند میزان تاب آوری جامعه ایرانی در برابر این بحران ها را به صورت هوشمند به سیاست گذاران نمایش داده و آن ها را نسبت به تغییرات افکار عمومی در این رابطه آگاه سازد.

Detection and Classification of Twitter Users' Opinions on Drought Crises in Iran Using Machine Learning Techniques

Drought crisis is one of the most significant environmental and social challenges in Iran, exerting widespread impacts on people's lives and the resilience of society. Identifying and analyzing the opinions of Persian-speaking Twitter users regarding this crisis can contribute to a better understanding of public opinion and support policymakers in their decision-making processes. This study aims to develop a model for detecting and categorizing user opinions related to drought crisis on the Twitter platform, utilizing machine learning and text mining methods to provide a more precise analysis of perspectives. At the outset of this study, Twitter content related to the drought crisis in Iran was systematically examined through the lens of Regulatory Focus Theory, enabling the categorization of user opinions within this theoretical framework. The dataset comprised 42,028 tweets collected over a one-year period using pertinent keywords associated with drought crises in Iran. A qualitative thematic analysis was subsequently performed on a representative sample of 2,300 tweets, which were meticulously labeled and classified. The thematic analysis revealed four principal categories of user opinions. In accordance with Regulatory Focus Theory, promotive tweets were further subdivided into Gain and Non-gain subcategories, while preventive tweets were classified into Non-losses and Losses subcategories. In the subsequent phase, a machine learning model employing logistic regression combined with word embedding vectors was developed to automatically predict and classify tweets into these defined categories. The dataset was processed and analyzed utilizing advanced text mining techniques and machine learning algorithms to ensure precise and robust categorization. The developed model effectively classified user opinions into four distinct categories related to the drought crisis and community resilience in Iran. It achieved an accuracy of 66.09% and an F1-score of 60%, reflecting a satisfactory level of performance in opinion classification. These results indicate that the model can reliably identify and categorize diverse user perspectives expressed on Twitter, thereby providing valuable insights to policymakers for a more informed understanding of public opinion dynamics. This study presents the proposed model as a novel contribution to the social sciences, particularly within the domain of drought and water crisis management, representing the first instance of framing this issue as a machine learning problem. The construction of a unique dataset, combined with the application of Regulatory Focus Theory to establish meaningful and contextually relevant categories, laid a robust foundation for the development of the machine learning framework. Integrating social science theories within the context of social media platforms facilitates more nuanced analyses aimed at understanding the underlying reasons and mechanisms driving changes in user behavior within these digital environments. Accordingly, this study applies Regulatory Focus Theory to extract and categorize user opinions expressed on Twitter concerning resilience to drought. Within this framework, Regulatory Focus Theory provides a foundational perspective for interpreting users’ motivations in producing and sharing content related to drought resilience on the Twitter platform. This research contributes to the field by developing a machine learning–based model for opinion detection and classification of Persian-language tweets addressing the drought crisis. Beyond advancing machine learning methodologies in this area, the study establishes a vital foundation for future research in the social sciences. The proposed model offers policymakers a valuable tool to intelligently evaluate community resilience and to gain deeper insights into public reactions during crises. Moreover, this study enriches the existing knowledge base in Persian opinion mining and its practical applications in crisis management. Future research can further enhance the model by incorporating larger datasets and employing more advanced deep learning techniques, as well as expanding its applicability to other environmental crises.

تبلیغات