Putri, Deannisa Syafira (2026) Forecasting Topik, Hashtag, dan Kata Berbasis LightGBM dengan Klasifikasi Interaksi Tweet Menggunakan MLP untuk Optimalisasi Marketing Campaign di X. Undergraduate thesis, UPN Veteran Jawa Timur.
|
Text (COVER)
Skripsi Cover.pdf Download (2MB) | Preview |
|
|
Text (BAB I)
Skripsi Bab 1.pdf Download (267kB) | Preview |
|
![[img]](https://repository.upnjatim.ac.id/style/images/fileicons/text.png) |
Text (BAB II)
Skripsi Bab 2.pdf Restricted to Repository staff only until 12 May 2028. Download (894kB) |
|
|
Text (BAB III)
Skripsi Bab 3.pdf Restricted to Repository staff only until 12 May 2028. Download (810kB) |
|
|
Text (BAB IV)
Skripsi Bab 4.pdf Restricted to Repository staff only until 12 May 2028. Download (7MB) |
|
|
Text (BAB V)
Skripsi Bab 5.pdf Download (238kB) | Preview |
|
|
Text (DAFTAR PUSTAKA)
Skripsi Daftar Pustaka.pdf Download (218kB) | Preview |
|
|
Text (LAMPIRAN)
Skripsi Lampiran.pdf Restricted to Repository staff only Download (340kB) |
Abstract
Social media, particularly platform X, exhibits rapid and widespread content dissemination patterns that create significant opportunities for data-driven marketing campaign optimization. This study develops a topic, hashtag, and word frequency forecasting system on platform X as the foundation for social media marketing campaign optimization, supported by a tweet interaction level classification model as supplementary analysis. Indonesian-language tweet data was collected through scraping over 18 months, yielding 3,110 clean records after preprocessing. Forecasting was performed using BERTopic for topic extraction and LightGBM optimized with Optuna as the forecasting model. The interaction level classification model was built by integrating IndoBERTweet, RoBERTa, and numerical metadata through a Multi-Layer Perceptron (MLP). Forecasting evaluation results demonstrate strong performance on hashtag data (RMSE 0.5470; MAE 0.3683; RMSSE 0.7554) and topic data (RMSE 1.5999; MAE 0.5783; RMSSE 0.7754), though performance remains suboptimal on word data (RMSE 3.2833; MAE 1.3928; RMSSE 1.4273). The classification model achieves accuracy of 0.8358, precision of 0.8351, recall of 0.8358, and F1-score of 0.8355. Integration of both models produces data-driven insights that support more targeted and effective content strategy development.
| Item Type: | Thesis (Undergraduate) | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Contributors: |
|
||||||||||||
| Subjects: | Q Science > QA Mathematics > QA76.6 Computer Programming Q Science > QA Mathematics > QA76.87 Neural computers |
||||||||||||
| Divisions: | Faculty of Computer Science > Departemen of Data Science | ||||||||||||
| Depositing User: | Deannisa Syafira Putri | ||||||||||||
| Date Deposited: | 20 May 2026 03:53 | ||||||||||||
| Last Modified: | 20 May 2026 04:04 | ||||||||||||
| URI: | https://repository.upnjatim.ac.id/id/eprint/51707 |
Actions (login required)
 |
View Item |