Prediction of LED Arrangement Illumination on Street Light Armature Using Random Forest Regression Method

LED's have a lifetime of up to 30 thousand hours. In addition, it has a level of effectiveness and efficiency of 80%-90%. The current problem is to determine the performance of the LED array arrangement using conventional methods so that the LED array arrangement must be made first and operated for testing. In this study, the LED array arrangement test determines each test sample's performance. Illumination prediction computations were performed on each sample using different test parameters. The prediction of illumination has a different predictive value in each test sample. The first and sixth samples have prediction accuracy below 95%. However, samples 2,3,4 and 5 have prediction accuracy above 95%. The interests of the current and temperature variables are 50% and 48%, while the voltage is only 3%. The value of R2, which is not by the research hypothesis, which is greater than 98%, is caused by the ability of the dependent variable, namely limited illumination. In addition, because the independent variables are less varied, they do not provide information on the dependent variable during the training data process. The low R2 value can also influence by the ability of the prediction test parameters formed.