We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Identifikasi Fitur untuk Prediksi Penerimaan Program Listrik Prabayar: Kasus di PLN Tahuna.
- Authors
Riduwan, Eko; Purwitasari, Diana; Raharjo, Agus Budi
- Abstract
Prepaid Electricity (LPB) provides benefits for government, electricity companies and customers. The State Electricity Company (PLN) has a marketing program to migrate from postpaid electricity to LPB. The achievement of Key Performance Indicators (KPI) for the Annual PLN LPB Marketing Program in 2021 was only 1,185 customers out of a target of 2,261 customers. This achievement provides an opportunity for improvement because the current marketing program has not optimized the use of data to determine customer prospects. This study proposes feature identification methods and scenarios for selecting the suitable Machine Learning algorithm to predict postpaid electricity customer acceptance of prepaid programs. Feature identification is made by measuring the Pearson correlation. The machine learning algorithm candidates selected are Logistic Regression, Support Vector Machines, Decision Tree, and Random Forest. The resulting models are evaluated using a confusion matrix to obtain the best model for the proposed case study. The research shows that the features of tariff, power frequency of late payment of electricity, average monthly electricity usage (kWh), and Regency significantly correlate with LPB revenue. The model with the Random Forest algorithm is the best model according to the research objectives, with the highest F1-Measure (95.17%).
- Subjects
MACHINE learning; RANDOM forest algorithms; SUPPORT vector machines; LATE payment; DECISION trees; PEARSON correlation (Statistics)
- Publication
Techno.com, 2022, Vol 21, Issue 3, p434
- ISSN
1412-2693
- Publication type
Article
- DOI
10.33633/tc.v21i3.6451