Data Analytics in Energy Systems: Applications of Keyword Extraction, Topic Modelling, and Text Classification in Text Mining

Öz

This study addresses the use of text mining and machine learning methods in the assessment of energy resources. The main objective is to reveal trends, themes and development areas associated with different types of renewable energy by analyzing large volumes of academic data in the literature. In this context, methods such as LDA (Latent Dirichlet Allocation), TF-IDF (Term Frequency-Inverse Document Frequency) and Naive Bayes were used to reveal the contextual structure of the texts and to examine how energy types are associated with each other. The dataset consists of the titles, keywords and abstracts of 25150 open access academic publications taken from the Web of Science database. These texts were vectorized by preprocessing (stop word removal, lemmatization, etc.) and then analyzed with various classification and topic modeling algorithms. In particular, TF-IDF method was used to identify prominent words, while LDA was used to identify possible thematic topics for each energy type. Naive Bayes was considered as a basic model for the classification of energy types. The results show that energy types such as solar, wind and biomass have a dominant place in the literature, whereas types such as wave, tidal and geothermal are underrepresented. In addition, text mining applications can identify not only existing trends but also areas where there are still research gaps. This study proposes a systematic data analytics approach that can guide future research in renewable energy systems. It also provides important contributions in terms of making the information density in the literature more meaningful for policy makers, academics and investors.

Anahtar Kelimeler

• Energy Types
• Text Mining
• LDA
• TF-IDF
• Naive Bayes

Enerji Sistemlerinde Veri Analizi: Metin Madenciliğinde Anahtar Kelime Çıkarma, Konu Modelleme ve Metin Sınıflandırmasının Uygulamaları

Öz

Bu çalışma, enerji kaynaklarının değerlendirilmesinde metin madenciliği ve makine öğrenmesi yöntemlerinin kullanımını ele almaktadır. Ana amaç, literatürdeki büyük miktarda akademik veriyi analiz ederek farklı yenilenebilir enerji türleriyle ilişkili eğilimleri, temaları ve geliştirme alanlarını ortaya çıkarmaktır. Bu bağlamda, metinlerin bağlamsal yapısını ortaya çıkarmak ve enerji türlerinin birbirleriyle nasıl ilişkili olduğunu incelemek için LDA (Latent Dirichlet Allocation), TF-IDF (Term Frequency-Inverse Document Frequency) ve Naive Bayes gibi yöntemler kullanılmıştır. Veri seti, Web of Science veri tabanından alınan 25150 açık erişimli akademik yayının başlıklarını, anahtar kelimelerini ve özetlerini içermektedir. Bu metinler ön işleme (durak kelime kaldırma, kök sözcük bulma vb.) ile vektörleştirilmiş ve daha sonra çeşitli sınıflandırma ve konu modelleme algoritmalarıyla analiz edilmiştir. Özellikle, TF-IDF yöntemi öne çıkan kelimeleri belirlemek için, LDA ise her enerji türü için olası tematik konuları belirlemek için kullanılmıştır. Naive Bayes, enerji türlerinin sınıflandırılması için temel bir model olarak ele alınmıştır. Sonuçlar, güneş, rüzgar ve biyokütle gibi enerji türlerinin literatürde baskın bir yere sahip olduğunu, dalga, gelgit ve jeotermal gibi türlerin ise yeterince temsil edilmediğini göstermektedir. Ayrıca, metin madenciliği uygulamaları yalnızca mevcut eğilimleri değil, aynı zamanda araştırma boşluklarının bulunduğu alanları da belirleyebilir. Bu çalışma, yenilenebilir enerji sistemlerinde gelecekteki araştırmalara rehberlik edebilecek sistematik bir veri analizi yaklaşımı önermektedir. Ayrıca, literatürdeki bilgi yoğunluğunu politika yapıcılar, akademisyenler ve yatırımcılar için daha anlamlı hale getirme açısından önemli katkılar sağlamaktadır.

Anahtar Kelimeler

• Enerji Türleri
• Metin Madenciliği
• LDA
• TF-IDF
• Naive Bayes

Kaynakça

1. [1] Reijinders L. Fuels for the future. Journal of Integrative Environmental Sciences. 2009; 6(4): 279–294.
2. [2] Bi H, Liu J, Wang L, Liu T, Zhang Z, Shen Q, Hayase S. Selective contact self-assembled molecules for high-performance perovskite solar cells. eScience. 2025; 100329.
3. [3] Ökten M. Turkey's energy outlook, goals, and future projections on the occasion of its centennial. Turkish Journal of Engineering Research and Education. 2023; 2(1): 52–60.
4. [4] Sovacool BK. The avian and wildlife costs of fossil fuels and nuclear power. Journal of Integrative Environmental Sciences. 2012; 9(4): 255–278.
5. [5] Zhang H, Li N, Gao S, Chen A, Qian Q, Kong Q, Xia BY, Hu G. Quenching-induced atom-stepped bimetallic sulfide heterointerface catalysts for industrial hydrogen generation. eScience. 2025; 100311.
6. [6] Hewitt NJ. Energy storage as a regional economy enabler. International Journal of Ambient Energy. 2016; 37(4): 329–330.
7. [7] Jiang H, Qiang M, Lin P. A topic modeling based bibliometric exploration of hydropower research. Renewable and Sustainable Energy Reviews. 2016; 57: 226–237.
8. [8] Benites-Lazaro LL, Giatti L, Giarolla A. Topic modeling method for analyzing social actor discourses on climate change, energy and food security. Energy Research & Social Science. 2018; 45: 318–330.

9. [9] Park C, Kim M. A study on the characteristics of academic topics related to renewable energy using the structural topic modeling and the weak signal concept. Energies. 2021; 14: 1497.
10. [10] Saheb T, Dehghani M, Saheb T. Artificial intelligence for sustainable energy: A contextual topic modeling and content analysis. Sustainable Computers: Informatics and Systems. 2022; 35: 100699.
11. [11] Choi H, Woo J. Investigating emerging hydrogen technology topics and comparing national level technological focus: Patent analysis using a structural topic model. Applied Energy. 2022; 313: 118898.
12. [12] Si Y. Co-authorship in energy justice studies: Assessing research collaboration through social network analysis and topic modeling. Energy Strategy Reviews. 2022; 41: 100859.
13. [13] Richarz J, Wegewitz S, Henn S, Müller D. Graph-based research field analysis by the use of natural language processing: An overview of German energy research. Technological Forecasting and Social Change. 2023; 186: 122139.
14. [14] Aydınoğlu AU, Erden Topal Y, Gökalp İ. Hidrojen teknolojileri araştırmalarında disiplinlerarasılık: Toplumsal bilimler yaklaşımı. Journal of Polytechnic. 2023; 26(1): 357–366.
15. [15] Şenyapar HND, Çetinkaya Ü, Bayındır R. Renewable energy incentives and future implications for Turkey: A comparative bibliometric analysis. Journal of Polytechnic. 2024; 27(1): 329–342.
16. [16] Van Eck NJ, Waltman L. Software survey: VOSviewer, a computer program for bibliometric mapping. Scientometrics. 2010; 84(2): 523–538.
17. [17] VOSviewer. Centre for Science and Technology Studies, Leiden University, The Netherlands; 2025.
18. [18] MATLAB. The MathWorks Inc., Natick, Massachusetts, USA; 2025.
19. [19] Blei DM, Ng AY, Jordan MI. Latent Dirichlet Allocation. Journal of Machine Learning Research. 2003; 3: 993–1022.
20. [20] Bhat MR, Kundroo MA, Tarray TA, Agarwal B. Deep LDA: A new way to topic model. Journal of Informatics Optimization Science. 2019; 41(3): 823–834.
21. [21] Tian C, Zhang J, Liu D, Wang Q, Lin S. Technological topic analysis of standard-essential patents based on the improved Latent Dirichlet Allocation (LDA) model. Technological Analysis and Strategic Management. 2022; 36(9): 2084–2099.
22. [22] Firoozeh N, Nazarenko A, Alizon F, Daille B. Keyword extraction: Issues and methods. Natural Language Engineering. 2020; 26: 259–291.
23. [23] Robertson S. Understanding inverse document frequency: On theoretical arguments for IDF. Journal of Documentation. 2004; 60: 503–520.
24. [24] Havrlant L, Kreinovich V. A simple probabilistic explanation of term frequency-inverse document frequency (TF-IDF) heuristic (and variations motivated by this explanation). International Journal of General Systems. 2017; 46(1): 27–36.
25. [25] Mendez M, Merayo MG, Nunez M. Design of hybrid machine learning and TF-IDF models to discard irrelevant reviews on public transport stations. Journal of Informatics and Telecommunication. 2025; 1–24.
26. [26] Yang H, Cui H, Tang H. A text classification algorithm based on feature weighting. AIP Conference Proceedings. 2017; 1864: 020026.
27. [27] Jiang L, Wang D, Cai Z, Jiang S, Yan X. Scaling up the accuracy of K-nearest-neighbour classifiers: A Naive-Bayes hybrid. International Journal of Computer Applications. 2009; 31(1): 36–43.
28. [28] Jiang L, Cai Z, Wang D. Improving Naive Bayes for classification. International Journal of Computer Applications. 2010; 32(3): 328–332.