http://repository.ipb.ac.id/handle/123456789/162418 2026-09-09T11:32:44Z http://repository.ipb.ac.id/handle/123456789/179809 Analisis Kinerja Arsitektur Transformer dan PatchTST pada Peramalan Harga Aset Safe haven Azahran, Muhammad Ryan Ketidakpastian ekonomi global mendorong meningkatnya minat terhadap aset safe haven sebagai instrumen lindung nilai, sekaligus memunculkan kebutuhan akan model peramalan harga yang andal. Penelitian ini menganalisis kinerja arsitektur Transformer dan PatchTST dalam meramalkan harga mingguan tiga aset safe haven terhadap rupiah, yaitu XAU/IDR, JPY/IDR, dan CHF/IDR, pada periode Januari 2016 hingga Desember 2025. Data dibagi menggunakan skema expanding window tiga lipatan, dengan dua lipatan pertama digunakan untuk pemilihan konfigurasi hyperparameter dan lipatan ketiga sebagai pengujian akhir tahun 2025. Penskalaan hanya diduga dari data latih pada setiap lipatan untuk mencegah kebocoran informasi. Konfigurasi terpilih adalah PatchTST untuk XAU/IDR serta Transformer untuk JPY/IDR dan CHF/IDR, dengan rata-rata MAPE pada tahap pemilihan sebesar 14,5%, 3,6%, dan 6,0% secara berturut-turut. Pada pengujian akhir tahun 2025, model terpilih memperoleh MAPE sebesar 19,7% untuk XAU/IDR, 5,4% untuk JPY/IDR, dan 8,5% untuk CHF/IDR, seluruhnya lebih rendah dibandingkan Naive Baseline yang mencatat MAPE 23,6%, 6,4%, dan 9,5%. Meskipun demikian, peramalan model terpilih cenderung mendatar dan belum menangkap dinamika harga pada periode dengan pergerakan ekstrem. Pada periode pengujian tahun 2025, konfigurasi model terpilih menghasilkan RMSE dan MAPE yang lebih rendah daripada Naive Baseline pada ketiga aset, namun Naive Baseline tetap merupakan tolok ukur penting yang perlu disertakan dalam evaluasi peramalan aset finansial.; Global economic uncertainty has increased interest in safe haven assets as hedging instruments, along with the need for reliable price forecasting models. This study analyzes the performance of Transformer and PatchTST architectures in forecasting the weekly prices of three safe haven assets against the rupiah, namely XAU/IDR, JPY/IDR, and CHF/IDR, over the period January 2016 to December 2025. The data were partitioned using a three-fold expanding window scheme, in which the first two folds were used for hyperparameter selection and the third fold served as the final test for the year 2025. Scaling parameters were estimated solely from the training data of each fold to prevent information leakage. The selected configurations were PatchTST for XAU/IDR and Transformer for JPY/IDR and CHF/IDR, with average MAPE values at the selection stage of 14.5%, 3.6%, and 6.0%, respectively. On the 2025 final test, the selected models obtained MAPE values of 19.7% for XAU/IDR, 5.4% for JPY/IDR, and 8.5% for CHF/IDR, all lower than the Naive Baseline, which recorded MAPE values of 23.6%, 6.4%, and 9.5%. Nevertheless, the forecasts of the selected models tended to be flat and did not capture price dynamics during periods of extreme movement. Over the 2025 test period, the selected model configurations produced lower RMSE and MAPE than the Naive Baseline for all three assets, yet the Naive Baseline remains an important benchmark that should be included in the evaluation of financial asset forecasting. 2026-01-01T00:00:00Z http://repository.ipb.ac.id/handle/123456789/179800 Identifikasi Peubah Penting terhadap Indeks Prestasi Mahasiswa D-IV Jalur SNBP dengan Random Forest dan EasyEnsemble Rahmawati, Elke Frida Indeks Prestasi (IP) Semester 1 merupakan indikator keberhasilan akademik awal mahasiswa yang dipengaruhi oleh berbagai faktor. Sementara itu, distribusi kategori IP pada mahasiswa D-IV jalur Seleksi Nasional Berdasarkan Prestasi (SNBP) cenderung tidak seimbang sehingga dapat menurunkan kemampuan model klasifikasi dalam mengenali kelas minoritas. Penelitian ini bertujuan membandingkan performa Random Forest standar dan Random Forest berbasis EasyEnsemble dalam mengklasifikasikan kategori IP Semester 1 serta mengidentifikasi peubah yang berkontribusi terhadap hasil klasifikasi menggunakan SHapley Additive exPlanations (SHAP). Data yang digunakan merupakan data primer mahasiswa D-IV angkatan 62 jalur SNBP IPB University tahun 2025. Evaluasi model dilakukan menggunakan repeated stratified hold out sebanyak lima kali ulangan dengan kombinasi hyperparameter ntree (50, 100, 150) dan mtry (3, 5, 7). Performa model dibandingkan menggunakan Balanced Accuracy, Macro F1-Score, dan Geometric Mean (G-Mean), sedangkan pengaruh hyperparameter dianalisis menggunakan ANOVA faktorial dua arah. Hasil penelitian menunjukkan bahwa Random Forest berbasis EasyEnsemble pada skema tiga kelas menghasilkan performa terbaik dengan peningkatan Balanced Accuracy dan G-Mean dibandingkan Random Forest standar. Analisis SHAP menunjukkan bahwa Rataan Pendidikan Kewarganegaraan, Rataan Bahasa Indonesia, Rataan Matematika, Rataan Prasyarat, dan Rataan Total merupakan peubah yang paling berkontribusi terhadap klasifikasi kategori IP Semester 1.; First-semester Grade Point Average (GPA) is an important indicator of students' early academic performance. However, the distribution of GPA categories among D-IV students admitted through the National Selection Based on Achievement (SNBP) pathway is imbalanced, which may reduce the ability of classification models to identify minority classes. This study compared the performance of the standard Random Forest and EasyEnsemble-based Random Forest models for classifying first-semester GPA categories and identified important variables contributing to the classification using SHapley Additive exPlanations (SHAP). The study used primary data from D-IV students of the 62nd cohort admitted through the SNBP pathway at IPB University in 2025. Model evaluation was conducted using repeated stratified holdout with five repetitions and combinations of ntree (50, 100, and 150) and mtry (3, 5, and 7). Model performance was assessed using Balanced Accuracy, Macro F1-Score, and Geometric Mean (G-Mean), while the effects of hyperparameters were evaluated using two-way ANOVA. The results showed that the EasyEnsemble-based Random Forest with the three-class scheme achieved the best performance by improving Balanced Accuracy and G-Mean compared with the standard Random Forest. SHAP analysis indicated that Civic Education average score, Indonesian Language average score, Mathematics average score, prerequisite subject average score, and overall report card average score were the most influential variables in classifying first-semester GPA categories. 2026-01-01T00:00:00Z http://repository.ipb.ac.id/handle/123456789/179639 Penerapan dan Evaluasi Fuzzy Geographically Weighted Clustering Berbasis Algoritma Genetika dalam Pemetaan Karakteristik Sosial Ekonomi di Provinsi Jawa Barat Suwandi, Farid Yafi Provinsi Jawa Barat masih menghadapi ketimpangan sosial ekonomi antarwilayah, tercermin dari Rasio Gini 0,428 (September 2024) dan Tingkat Pengangguran Terbuka (TPT) 6,75 persen (Agustus 2024) yang menjadi salah satu yang tertinggi secara nasional. Karena ketimpangan ini erat kaitannya dengan faktor geografis, metode klasterisasi konvensional yang mengabaikan lokasi wilayah dinilai kurang tepat. Penelitian ini menerapkan Fuzzy Geographically Weighted Clustering dengan optimasi inisialisasi centroid menggunakan Algoritma Genetika (FGWC-AG) untuk memetakan karakteristik sosial ekonomi 27 kabupaten/kota di Jawa Barat berdasarkan delapan indikator tahun 2024 (TPT, TPAK, UMK, laju pertumbuhan PDRB, pengeluaran per kapita, tingkat kemiskinan, rata-rata lama sekolah, dan angka harapan hidup). Algoritma Genetika digunakan untuk mencari kandidat centroid awal yang kemudian digunakan dalam iterasi FGWC hingga mencapai kriteria konvergensi. Hasil grid search menetapkan kombinasi parameter optimal pada c=2, a=0,9, dan ß=0,1. Konfigurasi ini menghasilkan struktur pengelompokan yang paling tegas dan kompak, ditunjukkan oleh nilai Partition Coefficient (PC) sebesar 0,636 dan Xie-Beni Index (XB) sebesar 0,367. Pada satu skenario eksekusi yang dilaporkan, FGWC-AG menghasilkan fungsi objektif dan indeks validitas yang secara numerik hampir sama dengan FGWC klasik. Hasil klasterisasi membagi wilayah menjadi dua kelompok: Klaster 1 (17 wilayah, agraris dan peri-urban di Priangan Timur serta Ciayumajakuning) dengan rata-rata UMK Rp2.747.777 dan kemiskinan 9,53 persen, serta Klaster 2 (10 wilayah metropolitan dan koridor industri) dengan rata-rata UMK jauh lebih tinggi (Rp4.429.222) dan kemiskinan lebih rendah (5,42 persen). Menariknya, TPT Klaster 2 justru lebih tinggi (7,74 persen) dibanding Klaster 1 (5,80 persen), yang mengindikasikan bahwa tingginya aktivitas ekonomi dan industrialisasi di kawasan perkotaan tidak serta-merta berbanding lurus dengan penyerapan tenaga kerja lokal secara optimal. Hasil pemetaan ini diharapkan menjadi bahan pertimbangan pemerintah dalam merumuskan kebijakan pembangunan yang lebih tepat sasaran sesuai karakteristik tiap daerah.; West Java Province continues to face socioeconomic disparities across regions, as reflected by a Gini Ratio of 0.428 (September 2024) and an Open Unemployment Rate (TPT) of 6.75 percent (August 2024), ranking among the highest nationally. Because these spatial disparities are closely linked to geographic factors, conventional clustering methods that ignore spatial locations are deemed inadequate. This study applies Fuzzy Geographically Weighted Clustering with centroid initialization optimization using a Genetic Algorithm (FGWC-AG) to map the socioeconomic characteristics of 27 regencies/cities in West Java based on eight 2024 indicators (TPT, LFPR, minimum wage, GRDP growth rate, per capita expenditure, poverty rate, mean years of schooling, and life expectancy). A Genetic Algorithm is used to search for initial centroid candidates, which are subsequently utilized in the FGWC iterations until convergence criteria are met. The grid-search results identifieed c=2, a=0.9, and ß=0.1 as the best configuration among the evaluated combinations. This configuration produced the most distinct and compact partition structure, as indicated by a Partition Coefficient (PC) of 0.636 and a Xie-Beni Index (XB) of 0.367. In one reported execution scenario, FGWC-AG produced an objective function and validity indices that were numerically nearly identical to classic FGWC. The final clustering divided the regions into two groups: Cluster 1 (17 agrarian and peri-urban regions in Eastern Priangan and Ciayumajakuning) with an average minimum wage of IDR 2,747,777 and a poverty rate of 9.53 percent, and Cluster 2 (10 metropolitan and industrial corridor regions) with a higher average minimum wage (IDR 4,429,222) and lower poverty rate (5.42 percent). Interestingly, the unemployment rate in Cluster 2 was higher (7.74 percent) than in Cluster 1 (5.80 percent), indicating that high economic activity and industrialization in urban areas do not necessarily correspond directly with optimal local labor absorption. These mapping results are expected to serve as a reference for policymakers in formulating targeted regional development strategies tailored to the unique characteristics of each region. 2026-01-01T00:00:00Z http://repository.ipb.ac.id/handle/123456789/179033 Penerapan SEM-PLS dalam Menganalisis Faktor-Faktor yang Memengaruhi Kepuasan dan Loyalitas Gen Z Pengguna Netflix Utami, Sabrina Afifah Putri Perkembangan teknologi digital, terutama layanan Subscription Video on Demand (SVOD) telah meningkatkan persaingan di antara penyedia layanan. Hal ini penting bagi perusahaan untuk memahami faktor-faktor yang memengaruhi kepuasan serta loyalitas pengguna. Penelitian ini bertujuan untuk menganalisis pengaruh variety of content, user experience, dan price fairness terhadap customer satisfaction, serta pengaruh customer satisfaction terhadap customer loyalty pada Gen Z pengguna Netflix. Data primer dikumpulkan melalui kuesioner daring sebanyak 245 responden Gen Z pengguna Netflix di wilayah Jabodetabek. Pengambilan sampel dilakukan menggunakan teknik purposive sampling. Data dianalisis menggunakan metode Structural Equation Modeling–Partial Least Squares (SEM-PLS). Hasil penelitian menunjukkan bahwa model pengukuran telah memenuhi persyaratan validitas dan reliabilitas, sedangkan model struktural menunjukkan kecocokan yang baik dengan nilai SRMR sebesar 0,065. Koefisien determinasi (R²) menunjukkan bahwa model mampu menjelaskan 63,9% varians pada customer satisfaction dan 46,5% varians pada customer loyalty. Pengujian hipotesis menunjukkan bahwa user experience, price fairness, dan variety of content berpengaruh positif dan signifikan terhadap customer satisfaction, sedangkan customer satisfaction berpengaruh positif dan signifikan terhadap customer loyalty. Customer satisfaction secara signifikan memediasi hubungan antara user experience, price fairness, dan variety of content dengan customer loyalty. User experience memiliki pengaruh paling kuat terhadap customer satisfaction, sedangkan customer satisfaction merupakan faktor yang paling kuat dalam menentukan customer loyalty. Temuan ini menegaskan pentingnya peningkatan layanan Netflix untuk meningkatkan customer satisfaction dan membangun customer loyalty di kalangan pengguna Gen Z.; The rapid growth of subscription-based video streaming services has intensified competition among service providers, making it essential for companies to understand the factors influencing customer satisfaction and customer loyalty. This study aims to analyze the effects of variety of content, user experience, and price fairness on customer satisfaction, as well as the effect of customer satisfaction on customer loyalty among Gen Z Netflix users. Primary data were collected through an online questionnaire distributed to 245 Gen Z Netflix users in the Greater Jakarta area (Jabodetabek) using a purposive sampling technique. The data were analyzed using the Structural Equation Modeling–Partial Least Squares (SEM-PLS) method. The results indicate that the measurement model satisfies the validity and reliability requirements, while the structural model demonstrates a good fit with an SRMR value of 0.065. The coefficients of determination (R²) show that the model explains 63.9% of the variance in customer satisfaction and 46.5% of the variance in customer loyalty. Hypothesis testing reveals that user experience, price fairness, and variety of content have positive and significant effects on customer satisfaction, while customer satisfaction has a positive and significant effect on customer loyalty. Furthermore, customer satisfaction significantly mediates the relationships between user experience, price fairness, and variety of content and customer loyalty. Among the antecedent variables, user experience has the strongest influence on customer satisfaction, while customer satisfaction is the strongest determinant of customer loyalty. These findings provide valuable insights for Netflix in improving its services to enhance customer satisfaction and foster customer loyalty among Gen Z users. 2026-01-01T00:00:00Z