Analysis and Prediction of Turbidity Level of Water Based on Ammonia Substance Using Random Forest and K-Nearest Neighbor

Abstract

Water is the primary need for human survival. The need for clean water is very important, both in a household scale and in an industrial scale. The clean water used and consumed by the community comes from river water processed and distributed by the Regional Drinking Water Company (PDAM). Water conditions before being treated and distributed contain various harmful substances if not purified, one of which is ammonia. Currently, with the development of information technology, especially in the field of machine learning and data analysis, the process of predicting the content of ammonia substances in water is becoming increasingly facilitated. Machine learning has provided a number of scientific prediction methods that can be used. In this research, the methods used were Random Forest Regression and K-Nearest Neighbor (KNN) methods are used to predict turbidity level of water in PDAM Surya Sembada Surabaya. This research aims to compare robbustness and accuracy of prediction models. The Random Forest method produced the best prediction error value of 0.0934, while the K-Nearest Neighbor (KNN) method produced the best prediction error value of 0.0918.