Pemanfaatan Machine Learning untuk Pengelompokan dan Prediksi Target Tambah Daya Listrik Pelanggan Prabayar (Studi Kasus : PT PLN ULP Watang Sawitto). (Indonesian)

Utilization of Machine Learning for Grouping and Predicting Targets for Adding Electricity to Prepaid Customers (Case Study: PT PLN ULP Watang Sawitto). (English)

The development of information system technology and science, especially in the field of marketing, makes business actors seek to increase their competitive advantage by mobilizing the resources owned by the company. Companies are required to innovate in managing their companies in order to survive in the world of competition. The ability to predict prepaid customers who have the potential to add electricity is one of the supporting strategies for the success of the marketing program for adding power to customers based on the characteristics of their electricity consumption. Based on this, this study proposes a prediction method for prepaid customers by utilizing clustering and classification algorithms. The data processed is household tariff prepaid customer data which features variable customer electrical power (VA), frequency of buying electricity tokens per year, total kWh usage per year, total rupiah token purchases per year, customer VA power difference, hours of operation, period days of purchase of electricity tokens, and history of adding electricity to customers. The grouping is done by applying the Kmeans algorithm. From these results, a prediction model is built according to the target of each cluster by utilizing two methods, Gradient Boosting and Artificial Neural Network. Evaluation of the best model predictions is carried out by applying three scenarios of the proportion of training data and test data, which are then measured using the accuracy matrix and Cohen Kappa. The experimental results produce four clusters based on the characteristics of their electricity consumption. Gradient Boosting gives the best results for all clusters, for cluster 1 it produces an AUC value of 0.784, cluster 2 produces an AUC value of 0.941, cluster 3 produces an AUC value of 0.884 and cluster 4 produces an AUC value of 0.903. [ABSTRACT FROM AUTHOR]

Perkembangan teknologi sistem informasi dan ilmu pengetahuan khususnya dalam bidang pemasaran membuat para pelaku usaha berupaya untuk meningkatkan competitive advantage mereka dengan mengerahkan sumber daya yang dimiliki oleh perusahaan. Perusahaan dituntut untuk berinovasi dalam mengelola perusahaannya agar dapat bertahan dalam dunia persaingan. Kemampuan untuk memprediksi pelanggan prabayar yang berpotensi tambah daya listrik merupakan salah satu strategi pendukung untuk keberhasilan program pemasaran tambah daya pelanggan berdasarkan karakteristik konsumsi listriknya. Berdasarkan hal tersebut, penelitian ini mengajukan metode prediksi pelanggan prabayar dengan memanfaatkan algoritma pengelompokan (Clustering) dan klasifikasi. Data yang diolah adalah data pelanggan prabayar tarif rumah tangga yang memiliki fitur variabel daya listrik pelanggan (VA), frekuensi beli token listrik, total pemakaian kWh, total rupiah pembelian token, selisih daya VA pelanggan, jam nyala, periode hari pembelian token listrik, dan riwayat tambah daya listrik pelanggan. Pengelompokan dilakukan dengan menerapkan algoritma K-means. Dari hasil tersebut, model prediksi dibangun sesuai target setiap klaster dengan memanfaatkan dua metode, Gradient Boosting dan Artificial Neural Network. Evaluasi prediksi model terbaik dilakukan dengan menerapkan tiga skenario proporsi data latih dan data uji, yang selanjutnya diukur menggunakan matrik akurasi dan Cohen Kappa. Hasil eksperimen menghasilkan empat klaster berdasarkan karakteristik konsumsi listriknya. Gradient Boosting memberikan hasil yang terbaik untuk semua klaster, untuk klaster 1 menghasilkan nilai AUC 0.784, klaster 2 menghasilkan nilai AUC 0.941, klaster 3 menghasilkan nilai 0.884 dan klaster 4 menghasilkan nilai AUC 0.903. [ABSTRACT FROM AUTHOR]

Copyright of Techno.com is the property of Universitas Dian Nuswantoro, Fakultas Ilmu Komputer and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)