Abstract:Keragaman harga, volume penjualan, frekuensi transaksi, dan persediaan produk Ulos menimbulkan kebutuhan akan strategi pemasaran yang berbeda sesuai karakteristik setiap produk. Analisis difokuskan pada perbandingan Single…
le Linkage, Complete Linkage, Average Linkage, dan Ward.D2 dalam Agglomerative Hierarchical Clustering sebagai dasar segmentasi produk di Galeri Ulos Sianipar Medan. Sebanyak 161 produk dianalisis menggunakan atribut harga jual, jumlah terjual, frekuensi transaksi, dan stok akhir. Perbedaan skala atribut ditangani melalui standardisasi Z-score, sedangkan kedekatan antarproduk dihitung menggunakan Euclidean Distance. Keempat metode linkage diterapkan untuk membentuk tiga klaster dan dievaluasi berdasarkan Silhouette Score, Davies-Bouldin Index, distribusi anggota, serta kemudahan interpretasi bisnis. Single Linkage menghasilkan Silhouette Score sebesar 0,556650 dan Davies-Bouldin Index sebesar 0,286346, tetapi membentuk distribusi yang sangat timpang, yaitu 1-2-158. Average Linkage dan Complete Linkage juga menghasilkan satu klaster dominan dengan distribusi masing-masing 1-147-13 dan 21-119-21. Ward.D2 memperoleh Silhouette Score sebesar 0,401513 dan Davies-Bouldin Index sebesar 0,799404, tetapi menghasilkan distribusi yang lebih proporsional, yaitu 21, 82, dan 58 produk. Dengan mempertimbangkan keseimbangan anggota, kejelasan centroid, dan relevansi hasil bagi pengambilan keputusan, Ward.D2 ditetapkan sebagai model operasional yang paling sesuai. Klaster yang terbentuk merepresentasikan produk ekonomis berperputaran cepat, produk reguler potensial, dan produk premium dengan penjualan selektif. Segmentasi tersebut mendukung pengendalian persediaan dan bundling, penguatan promosi digital dan edukasi produk, serta pemasaran premium melalui personal selling dan sistem pre-order.
Abstract:Abstract: This study aims to cluster provinces in Indonesia based on crime rates and community participation in security in 2024. The objective is to determine how effective the K-Means Clustering method is in analyzing…
this data. This study uses six crime indicators and five indicators of community participation at the provincial level. Before analysis, the data were cleaned and standardized using the Z-Score method. The clustering process employed the K-Means method with Euclidean distance. Cluster quality was evaluated using SSE and the Silhouette Score. The results show that the K-Means algorithm achieved convergence with a stable SSE value. However, determining the optimal number of clusters yielded different results between the Elbow method and the Silhouette Score. After considering stability and interpretability, three clusters were determined to be the optimal result. The clustering results indicate variations in characteristics across provinces regarding crime and community participation. The study also shows that cluster separation is better for crime data than for participation data. Thus, the K-Means Clustering method can be used as an initial mapping of regional security conditions in Indonesia. However, these results are also influenced by data characteristics and the sensitivity of centroid initialization.
Keywords: K-Means Clustering, Crime, Community Participation, Elbow Method, Silhouette Score.
Abstrak: Penelitian ini bertujuan untuk mengelompokkan provinsi di Indonesia berdasarkan tingkat kriminalitas dan partisipasi masyarakat dalam keamanan pada tahun 2024. Tujuannya adalah untuk mengetahui seberapa efektif metode K-Means Clustering dalam menganalisis data ini. Penelitian ini menggunakan enam indikator kriminalitas dan lima indikator partisipasi masyarakat di tingkat provinsi. Sebelum dianalisis, data tersebut dibersihkan dan distandardisasi menggunakan metode Z-Score. Proses pengelompokan menggunakan metode K-Means dengan jarak Euclidean. Kualitas cluster dievaluasi menggunakan SSE dan Silhouette Score. Dari hasil penelitian, algoritma K-Means dapat mencapai konvergensi dengan nilai SSE yang stabil. Namun, penentuan jumlah cluster optimal menunjukkan hasil yang berbeda antara metode Elbow dan Silhouette Score. Setelah mempertimbangkan stabilitas dan interpretasi, ditetapkan bahwa tiga cluster adalah hasil yang optimal. Hasil klasterisasi menunjukkan adanya variasi karakteristik antarprovinsi pada aspek kriminalitas dan partisipasi masyarakat. Hasil penelitian juga menunjukkan bahwa pemisahan cluster pada data kriminalitas lebih baik dibandingkan data partisipasi. Dengan demikian, metode K-Means Clustering dapat digunakan sebagai pemetaan awal kondisi keamanan wilayah di Indonesia. Namun, hasil ini juga dipengaruhi oleh karakteristik data dan sensitivitas inisialisasi centroid.
Kata kunci: K-Means Clustering, Kriminalitas, Partisipasi Masyarakat, Elbow Method, Silhouette Score.
Abstract:Violence against adult women is a social problem that still frequently occurs in Indonesia and shows an increase every year. Available data on violence cases are still in the form of raw data, making it difficult to use…
to determine the level of vulnerability in each region. Therefore, a method is needed that can process this data into more meaningful information through a classification process. This study aims to classify the level of violence against adult women in Indonesia using the Learning Vector Quantization (LVQ) algorithm. This data uses seven attributes, namely physical violence, psychological violence, sexual violence, exploitation, human trafficking, neglect, and others. The classification process is carried out into three classes, namely high, medium, and low. The test results show that the first data for the Aceh Province region is classified into class 1 (high) with a distance value of 111.7157645406. The second data for the North Sumatra Province region is also classified into class 1 (high) with a value of 114.41286640237. Meanwhile, the 3rd data for the West Sumatra Province region is classified into class 3 (low) with a value of 104.51753457. The results of the study indicate that the Learning Vector Quantization (LVQ) algorithm is able to group data on cases of violence against adult women based on their level of vulnerability so that it can be used as supporting information in decision-making and policy formulation for handling cases of violence against women.
Abstract:Equitable distribution of public facilities and infrastructure is a crucial challenge in regional governance in Aek Kuasan District, Asahan Regency. Currently, data inventories of educational, health, and religious facilities…
ities are still managed conventionally in tabular format or as a recapitulation of physical documents. This method makes it difficult for sub-district governments and stakeholders to identify spatial distribution patterns of facilities, often resulting in disparities in accessibility between villages, such as significant differences in facility density between densely populated residential areas and plantations. This study aims to design a web-based Geographic Information System (GIS) capable of automatically visualizing and analyzing public facility clustering. The method used is the K-Means Clustering algorithm with Euclidean distance measurement to partition the region into several clusters based on proximity of location coordinates. The frontend is built using React.js and the Leaflet map library for interactive visualization, while the backend uses Node.js to handle all the algorithm's computational logic, including distance calculations and centroid determination. Data storage is performed using a PostgreSQL relational database that stores location attributes in double-precision numeric format to ensure coordinate accuracy without relying on the database's spatial extensions. System testing results indicate that the application is capable of precisely mapping the locations of infrastructure facilities in seven villages/sub-districts in Aek Kuasan District. The implementation of the K-Means algorithm successfully classified areas into dense and sparse cluster zones, visualized through differences in marker colors on the digital map. This analytical information is expected to provide an empirical basis for the Aek Kuasan District Government in making more equitable and targeted development planning decisions.
Abstract:Abstract: Distribution of frozen food requires efficient route planning due to strict cold chain requirements and high operational costs. PT. Mitra Nasional Kualitas (Minaku) currently determines delivery routes based on…
drivers’ intuition, resulting in inefficiencies. This research applies Dijkstra’s algorithm integrated with real sales data to optimize distribution from Gresik to several cities in Java. The network is modeled as a weighted graph with distances as edge weights. Simulation on 15 cities with monthly demands exceeding 1,000 kg shows potential cost savings of about 16.2% when using optimized routing. Integration of shortest path algorithms with business data can significantly improve cost efficiency in cold chain logistics.
Keywords: Dijkstra algorithm, shortest path, logistics, frozen food, sales data.
Abstrak: Distribusi produk frozen food memerlukan perencanaan rute yang efisien karena tuntutan cold chain yang ketat dan biaya operasional yang tinggi. PT. Mitra Nasional Kualitas (Minaku) saat ini menentukan jalur distribusi berdasarkan intuisi sopir, sehingga berpotensi terjadi inefisiensi. Penelitian ini menerapkan Algoritma Dijkstra yang diintegrasikan dengan data penjualan riil untuk mengoptimasi distribusi dari Gresik ke beberapa kota di Pulau Jawa. Jaringan distribusi dimodelkan sebagai graf berbobot dengan jarak antar kota sebagai bobot sisi. Simulasi pada 15 kota dengan permintaan bulanan di atas 1.000 kg menunjukkan potensi penghematan biaya sekitar 16,2% dengan penggunaan rute optimal. Integrasi algoritma shortest path dengan data bisnis terbukti mampu meningkatkan efisiensi biaya dalam distribusi logistik cold chain.
Kata kunci: Algoritma Dijkstra, jalur terpendek, logistik, frozen food, data penjualan
Abstract:Oil palm is one of the plantation commodities that plays an important role in improving Indonesia's economy. PT Socfindo Kebun Aek Loba has a large amount of oil palm plant condition data; however, the data has not been…
optimally utilized to determine fertilizer dosage requirements. This study aims to classify oil palm fertilizer dosage requirements based on plant conditions using the K-Means algorithm and to design an application that supports the clustering process. The variables used in this study include plant age, tree height, number of fruit bunches, and number of fronds. The dataset consisted of 200 oil palm plant records. The clustering process was carried out by forming three clusters, namely low, medium, and high fertilizer dosage groups, using the K-Means method with Euclidean Distance calculations. The results showed that out of 200 plant data records processed, 92 data (46%) were classified into the low fertilizer dosage cluster, 54 data (27%) into the medium fertilizer dosage cluster, and 54 data (27%) into the high fertilizer dosage cluster. In addition, this study successfully developed an application using PHP and MySQL that is capable of managing data, performing clustering automatically, and presenting clustering results in the form of tables and charts. The developed application can assist PT Socfindo Kebun Aek Loba in obtaining information regarding fertilizer dosage requirements more quickly, effectively, and systematically, thereby supporting decision-making related to oil palm fertilization.
Abstract:Abstract: The massive growth of internet penetration in Indonesia has created competitive competition among internet service providers (ISPs), including Biznet Medan. However, the highly diverse and heterogeneous characteristics…
eristics of customers make one-size-fits-all marketing strategies ineffective and inefficient. Processing large customer data for segmentation is often hampered by the presence of outliers, such as extreme income variations, which can disrupt the clustering accuracy of conventional algorithms. This study aims to optimize the K-Means algorithm in segmenting Biznet Medan's target market to maintain accuracy despite outliers, allowing the company to design more targeted marketing strategies. To overcome K-Means' sensitivity to outliers, this study applies the Robust Scaling technique. This solution uses the Median and Interquartile Range (IQR) values instead of the mean, so that the cluster center points (centroids) remain stable and are not distorted by extreme data values. The research flow includes data collection of prospective Biznet Medan customers, data pre-processing (cleaning and scaling), feature selection (income, distance, and network requirements), and implementation of the K-Means model. Cluster quality evaluation is carried out by comparing the results of using StandardScaler and RobustScaler through the Silhouette Score and Davies Bouldin Index (DBI) metrics. The implementation successfully formed two main segments: Cluster 0 (Enterprise/Premium Customers) with high income characteristics and large bandwidth requirements (>200 Mbps), and Cluster 1 (Casual/Economical Customers) who dominate the market with lower network requirements (<50 Mbps). Optimization with Robust showed significant improvements in the evaluation metrics, where the Silhouette Score increased from 0.3133 to 0.6023, and the DBI value improved from 1.1421 to 0.8007. The implementation of Robust Scaling is effectively able to handle outlier problems in the K-Means algorithm, resulting in a more robust, efficient, and representative cluster structure. These results contribute to the processing of customer data for more objective strategic decision-making in ISP market expansion.
Keywords: k-means; clustering; outlier; robust scaling; customer segmentation.
Abstrak: Pertumbuhan internet di Indonesia yang masif telah menciptakan persaingan kompetitif di antara penyedia jasa internet (ISP), termasuk Biznet Medan. Namun, karakteristik pelanggan yang sangat beragam dan bersifat heterogen membuat strategi pemasaran satu untuk semua (one-size-fits-all) menjadi tidak efektif dan tidak efisien. Data calon pelanggan yang besar sering ditemui data yang hilang, perbedaan nilai data yang ekstrim sehingga pengolahan data calon pelanggan untuk segmentasi sering kali terkendala oleh adanya data pencilan (outlier) yang dapat mengganggu akurasi pengelompokan pada algoritma konvensional. Penelitian ini bertujuan untuk mengoptimasi algoritma K-Means dalam melakukan segmentasi target pasar Biznet Medan agar tetap akurat meskipun terdapat data outlier, sehingga perusahaan dapat merancang strategi pemasaran yang lebih tepat sasaran. Untuk mengatasi sensitivitas K-Means terhadap pencilan, penelitian ini menerapkan teknik Robust Scaling. Solusi ini menggunakan nilai Median dan Interquartile Range (IQR) sebagai pengganti rata-rata, sehingga titik pusat klaster (centroid) tetap stabil dan tidak terdistorsi oleh nilai data yang ekstrem. Alur penelitian meliputi pengumpulan data calon pelanggan Biznet Medan, pra-pemrosesan data (pembersihan dan scaling), pemilihan fitur (penghasilan, jarak, dan kebutuhan jaringan), serta implementasi model K-Means. Evaluasi kualitas klaster dilakukan dengan membandingkan hasil penggunaan StandardScaler dan RobustScaler melalui metrik Silhouette Score dan Davies Bouldin Index (DBI). Implementasi berhasil membentuk dua segmen utama: Cluster 0 (Pelanggan Enterprise/Premium) dengan karakteristik penghasilan tinggi dan kebutuhan bandwidth besar (>200 Mbps), serta Cluster 1 (Pelanggan Casual/Hemat) yang mendominasi pasar dengan kebutuhan jaringan lebih rendah (<50 Mbps). Optimasi dengan Robust menunjukkan peningkatan signifikan pada metrik evaluasi, di mana Silhouette Score meningkat dari 0,3133 menjadi 0,6023, dan nilai DBI membaik dari 1,1421 menjadi 0,8007. Penerapan Robust Scaling secara efektif mampu menangani masalah outlier pada algoritma K-Means, menghasilkan struktur klaster yang lebih kuat, efisien, dan representatif. Hasil ini memberikan kontribusi pada pengolahan data pelanggan untuk pengambilan keputusan strategis yang lebih objektif dalam ekspansi pasar ISP.
Kata kunci: K-Means; klasterisasi; data pencilan; Robust Scaling; Segmentasi Pelanggan.
Abstract:Abstract: This study aims to explore the pedagogical values within the Pilandok Takial Kial dance as an instrument for revitalizing youth ethics in Labuhanbatu Regency. In the digital era, teenagers face challenges of moral…
ral degradation and identity crisis that distance them from their local cultural roots. This study employs a qualitative approach with descriptive-analytical methods using Roland Barthes' Semiotic theory. Data were collected through textual observation of movements, literature studies, and in-depth interviews with dance maestros and traditional leaders. The results show that the Pilandok dance is not merely an aesthetic representation of a mousedeer’s movements but a system of signs containing strategic moral messages. Symbols of vigilance (eyes), resilience (takial-kial feet), and manners (worship gestures) are manifestations of the "civilized ingenuity" concept. The implications of this study emphasize the importance of integrating dance philosophy into character education curricula to shape a Generation Z that is intellectually bright yet still upholds Malay values of politeness.
Keywords: Pilandok Takial Kial Dance, Semiotics, Youth Ethics, Labuhanbatu, Cultural Revitalization.
Abstrak: Penelitian ini bertujuan untuk mengeksplorasi nilai-nilai pedagogis dalam Tari Pilandok Takial Kial sebagai instrumen revitalisasi etika remaja di Kabupaten Labuhanbatu. Di era digital, remaja menghadapi tantangan degradasi moral dan krisis identitas yang menjauhkan mereka dari akar budaya lokal. Penelitian ini menggunakan pendekatan kualitatif dengan metode deskriptif-analitis melalui teori Semiotika Roland Barthes. Data dikumpulkan melalui observasi tekstual gerak, studi literatur, serta wawancara mendalam dengan maestro tari dan tokoh adat. Hasil penelitian menunjukkan bahwa Tari Pilandok bukan sekadar representasi estetika gerak kancil, melainkan sebuah sistem tanda yang mengandung pesan moral strategis. Simbol kewaspadaan (mata), resiliensi (kaki takial-kial), dan adab (gestur sembah) merupakan manifestasi dari konsep "kecerdikan beradab". Implikasi penelitian ini menekankan pentingnya integrasi filosofi tari ke dalam kurikulum pendidikan karakter untuk membentuk Generasi Z yang cerdas secara intelektual namun tetap menjunjung tinggi nilai kesantunan Melayu.
Kata kunci: Tari Pilandok Takial Kial, Semiotika, Etika Remaja, Labuhanbatu, Revitalisasi Budaya.
Abstract:Rokan Hulu Regency possesses significant natural and cultural tourism potential; however, tourist visitation rates remain relatively low due to the lack of integrated destination information and concerns regarding environmental…
nmental sustainability. This study aims to design and develop SMART TOUR, a sequence-prediction–based tourism recommendation system utilizing a deep learning approach. The system is designed to learn tourist travel patterns based on their visitation history by employing a dual-input Long Short-Term Memory (LSTM) model that integrates location and activity data. To address the limitations of real-world data, a synthetic dataset was constructed in a realistic manner by considering tourist preferences, inter-location distances, visit duration, and the popularity of tourist attractions. Experimental results indicate that the proposed recommendation system achieves competitive performance, with a Top-1 Accuracy of 39.7%, Top-3 Accuracy of 76.2%, and Top-5 Accuracy reaching 92.4%. Evaluation was conducted using the Top-K Accuracy approach, which is considered most relevant in the context of recommendation systems as it reflects user behavior in selecting from multiple alternatives. This study demonstrates that integrating LSTM models with predictive techniques can enhance the quality of tourism recommendations and support the development of more environmentally friendly and sustainable tourism practices. SMART TOUR is expected to serve as an innovative solution to strengthen the tourism appeal of Rokan Hulu and empower local communities through the adoption of technology.
Abstract:Abstract: The oil palm plantation sector is a key pillar of the Asahan Regency economy. However, differences in regional characteristics lead to variations in land area, production, and productivity between sub-districts.…
. This study aims to cluster sub-districts in Asahan Regency based on oil palm production characteristics using the K-Medoids method. Data used comes from the Asahan Regency Agriculture Office, which covers 25 sub-districts with three main variables: mature area (ha), oil palm production (tons), and productivity (tons/ha/year). The clustering process was performed manually using Euclidean distance calculations, with the number of clusters set at three. The results indicate the formation of three main clusters: high, medium, and low production clusters. This grouping is expected to form the basis for more targeted and data-driven development planning for the oil palm plantation sector.
Keywords: K-Medoids, Clustering, Oil Palm, Plantation Production, Asahan Regency
Abstrak: Sektor perkebunan kelapa sawit merupakan salah satu penopang utama perekonomian Kabupaten Asahan. Namun, perbedaan karakteristik wilayah menyebabkan variasi luas lahan, produksi, dan produktivitas antar kecamatan. Penelitian ini bertujuan untuk mengelompokkan kecamatan di Kabupaten Asahan berdasarkan karakteristik produksi kelapa sawit menggunakan metode K-Medoids. Data yang digunakan berasal dari Dinas Pertanian Kabupaten Asahan yang mencakup 25 kecamatan dengan tiga variabel utama, yaitu luas tanaman menghasilkan (ha), produksi kelapa sawit (ton), dan produktivitas (ton/ha/tahun). Proses pengelompokan dilakukan secara manual menggunakan perhitungan jarak Euclidean dengan jumlah klaster ditetapkan sebanyak tiga. Hasil penelitian menunjukkan terbentuknya tiga klaster utama, yaitu klaster produksi tinggi, sedang, dan rendah. Pengelompokan ini diharapkan dapat menjadi dasar dalam perencanaan pembangunan sektor perkebunan kelapa sawit yang lebih terarah dan berbasis data.
Kata Kunci: K-Medoids, Clustering, Kelapa Sawit, Produksi Perkebunan, Kabupaten Asahan