Sulthan Kreshna, . (2025) PREDIKSI CUACA PER JAM UNTUK REKOMENDASI KENYAMANAN OLAHRAGA MENGGUNAKAN METODE BILSTM BERBASIS MOBILE. Skripsi thesis, Universitas Pembangunan Nasional Veteran Jakarta.
![[img]](http://repository.upnvj.ac.id/style/images/fileicons/text.png) |
Text
ABSTRAK.pdf Download (136kB) |
|
Text
AWAL.pdf Download (683kB) |
|
Text
BAB I.pdf Restricted to Repository UPNVJ Only Download (109kB) |
|
Text
BAB II.pdf Restricted to Repository UPNVJ Only Download (584kB) |
|
Text
BAB III.pdf Restricted to Repository UPNVJ Only Download (114kB) |
|
Text
BAB IV.pdf Restricted to Repository UPNVJ Only Download (3MB) |
|
Text
BAB V.pdf Download (92kB) |
|
Text
DAFTAR PUSTAKA.pdf Download (186kB) |
|
Text
DAFTAR RIWAYAT HIDUP.pdf Restricted to Repository UPNVJ Only Download (106kB) |
|
Text
LAMPIRAN.pdf Restricted to Repository UPNVJ Only Download (457kB) |
|
Text
HASIL PLAGIARISME.pdf Restricted to Repository staff only Download (26MB) |
|
Text
ARTIKEL.pdf Restricted to Repository staff only Download (960kB) |
Abstract
Indonesia's highly variable tropical climate has a significant impact on the transportation, agriculture, and industrial sectors, as well as outdoor activities such as sports. Accurate hourly weather forecasts for temperature, humidity, and rainfall are crucial to support human thermal comfort (measured via the Temperature Humidity Index/THI) and productivity. This study develops an hourly weather prediction model using a two-layer Bidirectional Long Short-Term Memory (BiLSTM) architecture and Yeo–Johnson transformation to address zero-inflation in rainfall data. Hourly historical data for Jakarta (January 1, 2021–December 31, 2023) from the Open-Meteo API was processed through cleaning, interpolation, and normalization (Yeo–Johnson + Min-Max Scaling). The time series with a 24-hour lookback and a 1-hour prediction horizon was divided into training (80%), validation (10%), and testing (10%) sets. The BiLSTM model (64→128 neurons, dropout 0.1–0.2, L2 regularization ≈1e−4) was compiled with the Adam optimizer and Huber loss, then optimized via random search and early stopping. The results show that the Yeo–Johnson transformation reduces rainfall skewness from 12.56 to 1.17. The model achieved a denormalized RMSE of 0.8787 °C (NSE 0.892) for temperature, 4.8653% (NSE 0.875) for humidity, and 0.7463 mm (NSE 0.368) for rainfall. The real-time Android-based API implementation successfully provided hourly predictions for sports comfort recommendations. The findings demonstrate the effectiveness of the BiLSTM and Yeo–Johnson combination in tropical weather prediction, with recommendations for variable enrichment and data periods for further research.
| Item Type: | Thesis (Skripsi) |
|---|---|
| Additional Information: | [No.Panggil: 2110511051] [Pembimbing: Indra Permana Solihin] [Penguji 1: Ridwan Raafi’udin] [Penguji 2: Nurul Afifah Arifuddin] |
| Uncontrolled Keywords: | BiLSTM; hourly weather prediction; Yeo–Johnson transformation; time series; |
| Subjects: | Q Science > Q Science (General) |
| Divisions: | Fakultas Ilmu Komputer > Program Studi Informatika (S1) |
| Depositing User: | SULTHAN KRESHNA MAHENDRA |
| Date Deposited: | 06 Aug 2025 06:53 |
| Last Modified: | 06 Aug 2025 06:53 |
| URI: | http://repository.upnvj.ac.id/id/eprint/38986 |
Actions (login required)
 |
View Item |