PENGEMBANGAN NEURAL NETWORK UNTUK PREDIKSI KUALITAS AIR

Abstract

Research on artificial intelligence to determine water quality has been widely developed as a human endeavor toimprove the quality of life. This study employs an artificial neural network (ANN) to determine the optimalclassification model for determining the safety of water. This study uses existing Kaggle generic datasets. Numerouspreprocesses were performed on the dataset starting from cleaning the data from missing values and outliers toequalizing the weights of each parameter with the min-max scaler. This study compares the accuracy of ANN modelin various scenarios constructed with 10, 15, 20, and 30 neurons. Scaled Conjugate Gradient is implemented as thelearning algorithm for developing the prediction model. The obtained results of the experiments vary betweenscenarios. Overall accuracy increases when the number of neurons is between 10 and 20, and decreases when thenumber of neurons is between 20 and 30.