Scientific Journal of Informatics Vol. 9, No. 1, May 2022 p-ISSN 2407-7658 http://journal.unnes.ac.id/nju/index.php/sji e-ISSN 2460-0040 84 | Scientific Journal of Informatics, Vol. 9, No. 1, May 2022 Combination of Backpropagation Neural Network and Particle Swarm Optimization for Prediction of Water Production in Municipal Waterworks Arif Agustyawan 1 *, Tri Ginanjar Laksana 2 , Ummi Athiyah 3 1,2 Informatics Department, Faculty of Informatics, Telkom Institute of Technology Purwokerto, Indonesia 3 Data Science Department, Faculty of Informatics, Telkom Institute of Technology Purwokerto, Indonesia Abstract. Purpose: As the population grows, the need for clean water also increases. Municipal Waterworks (PDAM) is an institution that regulates and manages the procurement of clean water for the community. So, the amount of water produced and distributed should be adjusted to the demand for water. Predictions on PDAM water production need to be done as planning and better preparation and facilitating and assisting in decision-making. Methods: The study used the Neural Network backpropagation algorithm combined with Particle Swarm Optimization (PSO) to predict the amount of water PDAM should produce. Backpropagation has a good ability to make predictions. But backpropagation has a weakness that causes it to get stuck at a local minimum. This is influenced by the determination of weights that are not optimal. In this study, PSO had a role in optimizing error values on the network to gain optimal weight. Result: This study obtained MSE values in the training and testing process of 0.00179 and 0.00081 from the combination model of backpropagation ANN and PSO. It is smaller than the ANN model without using an optimization algorithm. Novelty: The combination of JST backpropagation and PSO can improve predictions' accuracy and produce optimum weights. Keywords: PDAM Water Production, Backpropagation, ANN, PSO, Prediction Received April 2021 / Revised June 2021 / Accepted May 2022 This work is licensed under a Creative Commons Attribution 4.0 International License. INTRODUCTION Water is a necessity in everyday life [1]. All life in the world needs water [2]. Water is also a basic need for humans [3]. The existence of water on earth is limited. An accurate prediction is needed to anticipate water shortages in the future [2]. Municipal waterworks is a company that carries out service functions to produce drinking water and clean water for the community [1]. One of the problems in municipal waterworks is related to the availability of clean water production to meet customer needs precisely. Therefore, a study is required in order to provide a solution regarding the prediction of clean water that must be produced [4]. One of the prediction methods applied is Artificial Neural Network (ANN) [5]. Backpropagation architecture is one of several ANN architectures that can be used to study and analyze past data patterns more precisely to obtain a more accurate output [5]. In previous research conducted by Sihotang, dkk [6], the backpropagation ANN algorithm can predict the number of non-star hotel visitors with an accuracy of 88. Another study conducted by Hasan, dkk [7] was carried out forecasting the sales of Bottled Drinking Water (AMDK) for the 2019 period using backpropagation ANN obtained an MSE value of 0.00043743 and a MAPE value of 6.88%. In another study conducted Sutawinaya, dkk [8], the study was conducted to predict rainfall by comparing two ANN architectures, namely backpropagation ANN and Adaline ANN. The results showed that the RMSE value of backpropagation ANN testing was 0.0435, and Adaline ANN was 0.067. Another study was conducted by Nurkholiq, dkk [10] to predict the long-term demand for electrical energy. Two methods were compared, namely backpropagation ANN and fuzzy logic. This study * Corresponding author. Email addresses: 17102123@ittelkom-pwt.ac.id (Agustyawan), anjarlaksana@ittelkom-pwt.ac.id (Laksana), ummi@ittelkom-pwt.ac.id (Athiyah) DOI: 10.15294/sji.v9i1.29849