Azzah, Alyssa Amorita (2026) Prediksi Jumlah Penonton Program Televisi Stasiun Dangdut Menggunakan Metode Hybrid Generalized Autoregressive Conditional Heteroscedasticity dan Long Short-Term Memory. Undergraduate thesis, UPN Veteran Jawa Timur.
|
Text (Cover)
21083010057_Cover.pdf Download (627kB) | Preview |
|
|
Text (Bab 1)
21083010057_Bab 1.pdf Download (275kB) | Preview |
|
![[img]](https://repository.upnjatim.ac.id/style/images/fileicons/text.png) |
Text (Bab 2)
21083010057_Bab 2.pdf Restricted to Repository staff only until 21 May 2028. Download (591kB) |
|
|
Text (Bab 3)
21083010057_Bab 3.pdf Restricted to Repository staff only until 21 May 2028. Download (891kB) |
|
|
Text (Bab 4)
21083010057_Bab 4.pdf Restricted to Repository staff only until 21 May 2028. Download (1MB) |
|
|
Text (Bab 5)
21083010057_Bab 5.pdf Download (229kB) | Preview |
|
|
Text (Daftar Pustaka)
21083010057_Daftar Pustaka.pdf Download (235kB) | Preview |
|
|
Text (Lampiran)
21083010057_Lampiran.pdf Restricted to Repository staff only until 21 May 2028. Download (328kB) |
Abstract
Television stations often face difficulties in optimally planning program strategies and advertisement placements due to highly fluctuating viewership numbers. Daily viewership data of the Stasiun Dangdut program at JTV exhibits high volatility with a coefficient of variation of 33.35%, along with nonlinear patterns and long-term dependencies that are difficult to predict accurately using single conventional forecasting methods. To address this problem, this study develops a hybrid method integrating Generalized Autoregressive Conditional Heteroscedasticity (GARCH) and Long Short-Term Memory (LSTM) within a systematic analytical framework. The initial stage involved modeling the linear mean component using ARIMA(0,0,1) to extract residuals. Testing confirmed the presence of heteroscedasticity effects in the residuals, prompting the application of the GARCH(1,1) model with a Standardized Student's t distribution to model the dynamics of conditional volatility. The estimated conditional volatility from GARCH was then combined with historical viewership data, day of the week, and month as inputs to train the LSTM architecture, which is reliable in capturing nonlinear patterns and long-term dependencies. Evaluation of the hybrid GARCH-LSTM model yielded a Root Mean Squared Error (RMSE) of 9,713.24, a Mean Absolute Error (MAE) of 7,033.98, and a Mean Absolute Percentage Error (MAPE) of 20.51%, which is classified as reasonable according to Lewis (1982). Overall, the hybrid method proved effective in mapping general viewership movement trends, although limitations remain in replicating sudden extreme audience spikes caused by exogenous factors not represented in the historical data.
| Item Type: | Thesis (Undergraduate) | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Contributors: |
|
||||||||||||
| Subjects: | Q Science > Q Science (General) Q Science > QA Mathematics > QA76.6 Computer Programming |
||||||||||||
| Divisions: | Faculty of Computer Science > Departemen of Data Science | ||||||||||||
| Depositing User: | Alyssa Amorita Azzah | ||||||||||||
| Date Deposited: | 21 May 2026 07:38 | ||||||||||||
| Last Modified: | 21 May 2026 07:38 | ||||||||||||
| URI: | https://repository.upnjatim.ac.id/id/eprint/52073 |
Actions (login required)
 |
View Item |