{"title":"利用确定性因子优化人工神经网络模型","authors":"Syamsul Bachri, Kresno Sastro Bangun Utomo, Sumarmi Sumarmi, Mohammad Naufal Fathoni, Yulius Eka Aldianto","doi":"10.22146/MGI.57869","DOIUrl":null,"url":null,"abstract":"Kerawanan longsor di DAS Bendo termasuk dalam kerawanan kelas sedang hingga tinggi. Sampai dengan saat ini, pemetaan rawan longsor di DAS Bendo baru dilakukan pada skala pemetaan 1:250.000. Penelitian ini bertujuan untuk melakukan pemodelan pemetaan kerawanan longsor di DAS Bendo pada skala semi-detil. Metode yang digunakan dalam penelitian ini adalah optimalisasi model artificial neural network menggunakan certainty factor (C-ANN). Peta kerawanan dibangun berdasarkan faktor pengontrol tanah longsor yang berkorelasi positif terhadap kejadian longsor menggunakan Certainty Factor. Sedangkan pemodelan prediksi kerawanan menggunakan model ANN, khususnya arsitektur BPNN (back-propagation neural network). Hasil pemodelan menunjukkan bahwa model C-ANN (7 variabel independen) memiliki nilai AUC (0,916) lebih tinggi daripada model ANN (0,778). Faktor redundansi data, multikolinieritas data, dan proporsi kejadian longsor terhadap cakupan wilayah penelitian mengakibatkan ketidakpastian dalam data variabel independen. Melalui penelitian ini ditemukan hasil bahwa kondisi kerawanan longsor di DAS Bendo masuk kategori tinggi, khususnya pada lereng atas Gunung Ijen, Rante, dan Merapi. Landslide disaster in DAS Bendo is categorized as moderate to highly susceptible. Until today, landslide hazard mapping in DAS Bendo has been carried out with a scale 1:250.000. This study aimed to model landslide susceptibility mapping on a semi-detailed scale. The method used in this research was the integration of the Certainty Factor with Artificial Neural Network models (C-ANN).The development of susceptibility mapping based on factors that positively correlate to landslide events using Certainty Factor. While the susceptibility prediction model using the ANN model, specifically the BPNN (back-propagation neural network) architecture. Modelling results show that the C-ANN model (7 independent variables) has an AUC value (0.916) higher than the ANN model (0.778). Data redundancy factors, multicollinearity of data, and the proportion of landslide events to the study area's coverage resulted in uncertainty in the independent variable data. This research found that the Landslide hazard in the Bendo Watershed is in the high category, especially on the upper slopes of Mount Ijen, Rante, and Merapi.","PeriodicalId":55710,"journal":{"name":"Majalah Geografi Indonesia","volume":" ","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2021-02-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":"{\"title\":\"Optimalisasi Model Artificial Neural Network Menggunakan Certainty Factor (C-ANN) Untuk Pemetaan Kerawanan Tanah Longsor Skala Semi-Detil di DAS Bendo, Kabupaten Banyuwangi\",\"authors\":\"Syamsul Bachri, Kresno Sastro Bangun Utomo, Sumarmi Sumarmi, Mohammad Naufal Fathoni, Yulius Eka Aldianto\",\"doi\":\"10.22146/MGI.57869\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Kerawanan longsor di DAS Bendo termasuk dalam kerawanan kelas sedang hingga tinggi. Sampai dengan saat ini, pemetaan rawan longsor di DAS Bendo baru dilakukan pada skala pemetaan 1:250.000. Penelitian ini bertujuan untuk melakukan pemodelan pemetaan kerawanan longsor di DAS Bendo pada skala semi-detil. Metode yang digunakan dalam penelitian ini adalah optimalisasi model artificial neural network menggunakan certainty factor (C-ANN). Peta kerawanan dibangun berdasarkan faktor pengontrol tanah longsor yang berkorelasi positif terhadap kejadian longsor menggunakan Certainty Factor. Sedangkan pemodelan prediksi kerawanan menggunakan model ANN, khususnya arsitektur BPNN (back-propagation neural network). Hasil pemodelan menunjukkan bahwa model C-ANN (7 variabel independen) memiliki nilai AUC (0,916) lebih tinggi daripada model ANN (0,778). Faktor redundansi data, multikolinieritas data, dan proporsi kejadian longsor terhadap cakupan wilayah penelitian mengakibatkan ketidakpastian dalam data variabel independen. Melalui penelitian ini ditemukan hasil bahwa kondisi kerawanan longsor di DAS Bendo masuk kategori tinggi, khususnya pada lereng atas Gunung Ijen, Rante, dan Merapi. Landslide disaster in DAS Bendo is categorized as moderate to highly susceptible. Until today, landslide hazard mapping in DAS Bendo has been carried out with a scale 1:250.000. This study aimed to model landslide susceptibility mapping on a semi-detailed scale. The method used in this research was the integration of the Certainty Factor with Artificial Neural Network models (C-ANN).The development of susceptibility mapping based on factors that positively correlate to landslide events using Certainty Factor. While the susceptibility prediction model using the ANN model, specifically the BPNN (back-propagation neural network) architecture. Modelling results show that the C-ANN model (7 independent variables) has an AUC value (0.916) higher than the ANN model (0.778). Data redundancy factors, multicollinearity of data, and the proportion of landslide events to the study area's coverage resulted in uncertainty in the independent variable data. This research found that the Landslide hazard in the Bendo Watershed is in the high category, especially on the upper slopes of Mount Ijen, Rante, and Merapi.\",\"PeriodicalId\":55710,\"journal\":{\"name\":\"Majalah Geografi Indonesia\",\"volume\":\" \",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2021-02-25\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"2\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Majalah Geografi Indonesia\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.22146/MGI.57869\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Majalah Geografi Indonesia","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.22146/MGI.57869","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 2
摘要
DAS Bendo的纵向友谊包括在教室的亲戚中,达到了很高的水平。到目前为止,DAS Bendo码头线的测绘刚刚完成,比例尺为1:250000。本研究旨在对DAS Bendo的纵向客户映射进行半详细的建模。本研究中使用的方法是使用确定性因子(C-ANN)对人工神经网络模型进行优化。自定义地图是基于纵向土地控制因素构建的,这些因素与使用确定性因素的纵向事件呈正相关。而客户预测建模则采用了人工神经网络模型,尤其是反向传播神经网络架构。建模结果表明,模型C-ANN(7个自变量)的AUC值(0.916)高于模型ANN(0.778)。数据冗余因素、数据的多重共线性以及纵向事件占研究区域覆盖范围的比例导致自变量数据的不确定性。通过研究发现,DAS Bendo的纵向亲缘关系状况属于较高的类别,尤其是在Ijen山、Rante山和Merapi山以上的斜坡上。[UNK]DAS Bendo的滑坡灾害属于中度至高度易感灾害。直到今天,DAS Bendo的滑坡灾害测绘工作已经完成,比例尺为1:250.000。本研究旨在建立半详细规模的滑坡易发性绘图模型。本研究中使用的方法是将确定性因子与人工神经网络模型(C-ANN)相结合。使用确定性因子开发基于与滑坡事件正相关因素的易感性映射。而易感性预测模型采用的是ANN模型,特别是BPNN[UNK](反向传播神经网络)[UNK]架构。建模结果表明,C-ANN模型(7个自变量s)的AUC值(0.916)高于ANN模型(0.778)。数据冗余因素、数据的多重共线性以及滑坡事件占研究区域覆盖范围的比例导致自变量数据的不确定性。本研究发现,Bendo流域的滑坡灾害属于高级别,尤其是在Ijen山、Rante和Merapi的上坡。
Optimalisasi Model Artificial Neural Network Menggunakan Certainty Factor (C-ANN) Untuk Pemetaan Kerawanan Tanah Longsor Skala Semi-Detil di DAS Bendo, Kabupaten Banyuwangi
Kerawanan longsor di DAS Bendo termasuk dalam kerawanan kelas sedang hingga tinggi. Sampai dengan saat ini, pemetaan rawan longsor di DAS Bendo baru dilakukan pada skala pemetaan 1:250.000. Penelitian ini bertujuan untuk melakukan pemodelan pemetaan kerawanan longsor di DAS Bendo pada skala semi-detil. Metode yang digunakan dalam penelitian ini adalah optimalisasi model artificial neural network menggunakan certainty factor (C-ANN). Peta kerawanan dibangun berdasarkan faktor pengontrol tanah longsor yang berkorelasi positif terhadap kejadian longsor menggunakan Certainty Factor. Sedangkan pemodelan prediksi kerawanan menggunakan model ANN, khususnya arsitektur BPNN (back-propagation neural network). Hasil pemodelan menunjukkan bahwa model C-ANN (7 variabel independen) memiliki nilai AUC (0,916) lebih tinggi daripada model ANN (0,778). Faktor redundansi data, multikolinieritas data, dan proporsi kejadian longsor terhadap cakupan wilayah penelitian mengakibatkan ketidakpastian dalam data variabel independen. Melalui penelitian ini ditemukan hasil bahwa kondisi kerawanan longsor di DAS Bendo masuk kategori tinggi, khususnya pada lereng atas Gunung Ijen, Rante, dan Merapi. Landslide disaster in DAS Bendo is categorized as moderate to highly susceptible. Until today, landslide hazard mapping in DAS Bendo has been carried out with a scale 1:250.000. This study aimed to model landslide susceptibility mapping on a semi-detailed scale. The method used in this research was the integration of the Certainty Factor with Artificial Neural Network models (C-ANN).The development of susceptibility mapping based on factors that positively correlate to landslide events using Certainty Factor. While the susceptibility prediction model using the ANN model, specifically the BPNN (back-propagation neural network) architecture. Modelling results show that the C-ANN model (7 independent variables) has an AUC value (0.916) higher than the ANN model (0.778). Data redundancy factors, multicollinearity of data, and the proportion of landslide events to the study area's coverage resulted in uncertainty in the independent variable data. This research found that the Landslide hazard in the Bendo Watershed is in the high category, especially on the upper slopes of Mount Ijen, Rante, and Merapi.