K Fold Cross Validation Analysis for Electricity Meter Classification at PLN Lhoksukon Using K-NN and SVM Methods

Electricity consumption continues to increase every year in line with the increase in national economic growth. Predicting current electricity demand is important to understand the overall electricity supplied to each region. The problem is that currently, the electricity supply in various areas of Lhoksukon has not matched the needs of the community. In addition, problems can arise if the power generated is less than the load power requirements, causing energy shortages in an area. To find out whether the electricity provided is appropriate or not, a classification using Supervised Learning method is used. After classification, we will use K-fold Cross Validation to measure how good the accuracy is between the methods. This study will use 200 electricity meter data consisting of 150 test data and 50 training data with a composition of 75%: 25%. The testing process where the data process that has been divided is then carried out in the testing process where the data process is obtained from manual calculations. So that in this study get results in the form of the K-NN method with 99.3% accuracy, 100% precision, 99.29% recall and the SVM method with 94.00% accuracy, 94.00% precision, 100% recall. And to find out how well the performance of the method is based on Supervised Learning method, it will be checked using K-Fold Cross Validation with the results of K-NN 99.53% and SVM 96.00%, with the conclusion that the K-Nearest Neighbor method has a better accuracy rate.