Prediction of citation count using abstracts
(2025) STAN40 20251Department of Statistics
- Abstract
- An LSTM-based deep learning model, as well as a Lasso and linear regression model, were implemented to try and predict citation counts based on the abstracts. In an attempt to try and model the relationship between the citation count and the abstracts, specifically in the field of chemistry. A comparison between the Lasso model and the deep learning model was done, in order to weigh the interpretability of Lasso against the complexity of the deep learning model. The architecture of the deep learning model included layers such as Bi-LSTM, LSTM, attention, and an embedding layer using GloVe pre-trained embeddings. The results, surprisingly, were that the deep learning model and the Lasso model performed similarly. Lasso also had the... (More)
- An LSTM-based deep learning model, as well as a Lasso and linear regression model, were implemented to try and predict citation counts based on the abstracts. In an attempt to try and model the relationship between the citation count and the abstracts, specifically in the field of chemistry. A comparison between the Lasso model and the deep learning model was done, in order to weigh the interpretability of Lasso against the complexity of the deep learning model. The architecture of the deep learning model included layers such as Bi-LSTM, LSTM, attention, and an embedding layer using GloVe pre-trained embeddings. The results, surprisingly, were that the deep learning model and the Lasso model performed similarly. Lasso also had the advantage of having interpretable coefficients, which made it possible to find the words in the abstracts contributing the most to a high citation count. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9210963
- author
- Kornö, Lotta LU
- supervisor
- organization
- course
- STAN40 20251
- year
- 2025
- type
- H1 - Master's Degree (One Year)
- subject
- language
- English
- id
- 9210963
- date added to LUP
- 2025-09-19 10:10:56
- date last changed
- 2025-09-19 10:10:56
@misc{9210963, abstract = {{An LSTM-based deep learning model, as well as a Lasso and linear regression model, were implemented to try and predict citation counts based on the abstracts. In an attempt to try and model the relationship between the citation count and the abstracts, specifically in the field of chemistry. A comparison between the Lasso model and the deep learning model was done, in order to weigh the interpretability of Lasso against the complexity of the deep learning model. The architecture of the deep learning model included layers such as Bi-LSTM, LSTM, attention, and an embedding layer using GloVe pre-trained embeddings. The results, surprisingly, were that the deep learning model and the Lasso model performed similarly. Lasso also had the advantage of having interpretable coefficients, which made it possible to find the words in the abstracts contributing the most to a high citation count.}}, author = {{Kornö, Lotta}}, language = {{eng}}, note = {{Student Paper}}, title = {{Prediction of citation count using abstracts}}, year = {{2025}}, }