Cooling Load Energy Performance of Residential Building: Machine Learning-Cluster K-Nearest Neighbor CKNN (Part I)

In this paper, we perform energy analysis in assessing the cooling load of building shapes system based on the Cluster K-Nearest Neighbor (CKNN) method for classification. The system is implemented and simulated in Anaconda, and its performance is tested on real dataset that contains 8 features with 768 instances to classify the cooling load magnitude into 8 classes (8 target name labels). Classes were created separately based on the magnitude of captured cooling load. Various training and test sizes are used to compare the performance measurement of the cooling load energy to predict the class label description, when k, the nearest neighbor changes from 1 to 19. The simulation results show that the accuracy depends on both independent parameters, the test size and k-neighbors, which gives better training and no bad test accuracy rate with slightly lower in the range of [55%–100%] and [35%–75%], respectively. When k is in the interval [3–6], the training accuracy is approximately equal to [0.80 \( \pm \) 0.08], and the test accuracy, except for the test size = 10%, 40% and 50% is about [0.65 \( \pm \) 0.10]. Each class shows three (03) regions: 1) Region (I), in the range of k [1–6], where the accuracy increases or decreases when k increases; 2) Region (II), in the interval of k [6–10], the accuracy decreases as k increases, and 3) finally, in the region (III), where the accuracy remains approximately constant when k increases from 10 to 19. In this investigation, the prediction of the cooling load magnitude under different classes maybe optimized in the interval of k [4, 6] by combining the test size in the range of [10%–50%]. The present proposed methodology can serve as a platform how to utilize the machine learning techniques for measurement and verification of the energy cooling load, by defining a boundary of analysis based on the independent parameters such as k-neighbors and test size and makes use of all data recorded across the facility for the purposes of maximizing the accuracy.