Please use this identifier to cite or link to this item:
http://repository.ipb.ac.id/handle/123456789/169320Full metadata record
| DC Field | Value | Language |
|---|---|---|
| dc.contributor.advisor | Giri, Endang Purnama | |
| dc.contributor.advisor | Hasibuan, Lailan Sahrina | |
| dc.contributor.author | Lazuardi, Andhika Rafi | |
| dc.date.accessioned | 2025-08-14T16:35:41Z | |
| dc.date.available | 2025-08-14T16:35:41Z | |
| dc.date.issued | 2025 | |
| dc.identifier.uri | http://repository.ipb.ac.id/handle/123456789/169320 | |
| dc.description.abstract | Penelitian ini menganalisis dampak teknik imputasi terhadap performa prediksi kematian babi akibat wabah African Swine Fever (ASF), penyakit yang menular dan mematikan pada babi, di Sumatera Utara menggunakan model Deep Learning. Data diperoleh dari Balai Veteriner Medan dan disusun dalam format harian. Tahap praproses meliputi pengecekan dan pengisian missing value menggunakan teknik imputasi linear, moving average, PCHIP, dan Bayesian Gaussian Approach. Data dimodelkan menggunakan LSTM dan GRU, serta dievaluasi menggunakan metrik NRMSE, Pearson Correlation, dan MAPE. Secara kuantitatif, pada data harian diperoleh NRMSE = 0,888, Pearson = 0,9706, dan MAPE = 83,88%; setelah agregasi menjadi mingguan NRMSE meningkat menjadi 0,1695, Pearson turun menjadi 0,8852, dan MAPE melonjak menjadi 696,6914%. Hasil menunjukkan bahwa pemilihan teknik imputasi berpengaruh signifikan terhadap kualitas prediksi, di mana kombinasi antara imputasi PCHIP dengan model GRU pada data harian dapat memberikan performa terbaik dengan stabil dan akurat. Agregasi ke level mingguan menunjukkan penurunan performa, yang kemungkinan disebabkan oleh penurunan jumlah sampel data setelah agregasi. Temuan ini menegaskan pentingnya pemilihan metode imputasi yang tepat pada deret waktu dengan tingkat proporsi nilai hilang yang tinggi, serta bahwa keputusan agregasi harus dipertimbangkan sebagai trade-off dalam upaya menghasilkan prediksi yang andal dalam mendukung pemantauan dan pengambilan keputusan epidemiologis. | |
| dc.description.abstract | This study examines the impact of imputation techniques on predicting swine mortality due to African Swine Fever (ASF), a contagious and often fatal disease in domestic and wild pigs in North Sumatra using deep learning. Case data were obtained from Balai Veteriner Medan and organized in a daily format. Preprocessing included checking and imputing missing values using linear interpolation, moving average, PCHIP, and a Bayesian Gaussian approach. Data were modeled with LSTM and GRU and evaluated using NRMSE, Pearson correlation, and MAPE. Quantitatively, the daily results were NRMSE = 0,0888, Pearson = 0,9706, and MAPE = 83,88%; after aggregation to weekly data, NRMSE rose to 0,1695, Pearson fell to 0,8852, and MAPE surged to 696,69%. Results indicate that the choice of imputation significantly affects prediction quality: the PCHIP–GRU combination on daily data yielded the best, most stable and accurate performance. Aggregation to the weekly level reduced performance, likely due to the decreased number of observations. These findings underscore the importance of selecting appropriate imputation methods for time series with a high proportion of missing values, and that the decision to aggregate should be treated as a trade-off between noise reduction and loss of temporal resolution when producing reliable predictions for epidemiological monitoring and decision making. | |
| dc.description.sponsorship | ||
| dc.language.iso | id | |
| dc.publisher | IPB University | id |
| dc.title | Analisis Dampak Teknik Imputasi terhadap Performa Prediksi Kematian Ternak akibat African Swine Fever menggunakan Model LSTM dan GRU | id |
| dc.title.alternative | Impact Analysis of Imputation Techniques on Livestock Mortality Prediction Performance due to African Swine Fever Using LSTM and GRU Models. | |
| dc.type | Skripsi | |
| dc.subject.keyword | African swine fever | id |
| dc.subject.keyword | Time Series | id |
| dc.subject.keyword | deep learning | id |
| dc.subject.keyword | pig mortality | id |
| dc.subject.keyword | missing data imputation | id |
| Appears in Collections: | UT - Computer Science | |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| cover_G6401211139_9966cbc8acbd4205b9bcbe819cfde0a6.pdf | Cover | 2.32 MB | Adobe PDF | View/Open |
| fulltext_G6401211139_4ea37c172afc4c63a6685afd8d4160a2.pdf Restricted Access | Fulltext | 3.5 MB | Adobe PDF | View/Open |
| lampiran_G6401211139_669b983ba2c94aa8b429ded08e7c378d.pdf Restricted Access | Lampiran | 2.39 MB | Adobe PDF | View/Open |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.