Optimal Parameter Selection of Latent Dirichlet Allocation to Determine the Emerging Topics in Hydrology Domain
2nd International Conference on Computing, IoT and Data Analytics, ICCIDA 2023, Ciudad Real, İspanya, 20 - 21 Temmuz 2023, cilt.1145 SCI, ss.32-42, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Cilt numarası: 1145 SCI
- Doi Numarası: 10.1007/978-3-031-53717-2_4
- Basıldığı Şehir: Ciudad Real
- Basıldığı Ülke: İspanya
- Sayfa Sayıları: ss.32-42
- Anahtar Kelimeler: Hyperparameters, Latent Dirichlet Allocation, Text Analytics, Topics in Hydrology
- Dokuz Eylül Üniversitesi Adresli: Hayır
Özet
In the new digital age, the determination of emerging topics has become a central issue for academia. Latent Dirichlet Allocation (LDA) method, a key mechanism for determining trends, has long been a method of great interest in a wide range of fields. However, the probabilistic structure leads to a serious effect on the score not only by changing the parameters but also from trial to trial for the fixed parameters. This study describes the implementation of the LDA method for exploring trend topics in the hydrology domain. Several parameters like the portion of the corpus, the number of topics as well as hyperparameters of the LDA method, α and β, have been considered. The emerging topics of the field have been obtained using the parameters which attained the highest mean coherence score.