<![CDATA[Ozone tropospher (O3) is one of the pollutants in the environment of Mataram City, Lombok, NTB, Indonesia. Based on the data obtained from the Agency of Environment and Forestry of West Nusa Tenggara Province, ozone pollutant concentrations in Mataram City have changed unpredictably. One time pollutant concentrations increase and then decrease, but then quickly increase again significantly. Therefore, the concentrations of ozone pollutant must be monitored because its presence at certain levels can cause various negative effects human health and the environment. Changes in ozone pollutant concentrations can be identified by carrying out a method of predicting ozone pollutant levels so that a decision can be taken to prevent the negative impact of the pollutant. In this research, a backpropagation artificial neural network is used to find the model prediction of the concentration of ozone in Mataram City. The input variables that are used in this network are air temperature (x_1 ), wind direction (x_2 ), wind speed (x_3 ), humidity (x_4 ), solar radiation (x_5 ), concentration of NO2 (x_6 ), the concentration of SO2 (x_7 ) and the concentration of O3 a day before (x_8 ) for the period of 6 July 2018 to 31 May 2019. The method in this study was to conduct trial and error on 60 different combinations of network architectures and parameters. Then all the network architectures performance will be compared based on the RMSE, MAPE and R2 indicators. Based on this research, the best neural network model to predict the concentration of ozone pollutant in Mataram City is the network with architecture 8-20-1, with logsig-purelin activation function and trainlm learning function. The performance of the training model is RMSE=0.011, MAPE = 1,043 % and R^2=0,9566. Meanwhile, the performance of the testing model is RMSE=0.001, MAPE = 0.749 % and R^2=0.497]]>
