Energy Consumption Level Classification using Stepwise Dynamic Nearest Neighbor Algorithm

Karabas D., BİRANT D., Taser P. Y.

7th International Symposium on Innovative Approaches in Smart Technologies, ISAS 2023, İstanbul, Türkiye, 23 - 25 Kasım 2023, (Tam Metin Bildiri) identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/isas60782.2023.10391690
  • Basıldığı Şehir: İstanbul
  • Basıldığı Ülke: Türkiye
  • Anahtar Kelimeler: classification, energy consumption, ensemble learning, k-nearest neighbor, machine learning
  • Dokuz Eylül Üniversitesi Adresli: Evet

Özet

Predicting energy consumption is a crucial element of responsible energy management as it guarantees the effective allocation of resources, the mitigation of costs, and the transition to sustainable energy sources. Machine learning (ML) has become increasingly prominent in recent times due to its capacity to generate exceptionally precise predictions through the utilization of diverse data sources. However, with a restricted focus on classification studies, the existing studies in the literature generally emphasize regression-type works that try to predict the numerical energy value based on historical data. To bridge this gap, a stepwise dynamic nearest neighbor (SDNN), which is an improved k-nearest neighbor (KNN) algorithm, is applied to three real-world datasets for classifying energy consumption levels of commercial center-type consumers. To demonstrate the effectiveness of the applied approach, the SDNN algorithm was compared with the traditional KNN, two Multilayer Perceptron models (MLP-1, MLP-2), Support Vector Machine (SVC), and Random Forest (RF) algorithms. It was observed from the experimental results that the SDNN algorithm achieved the best energy consumption level prediction among the others, with an average accuracy rate of 92.82%.