PENERAPAN MODEL ARIMA DALAM PENDEKATAN MACHINE LEARNING UNTUK PERAMALAN KONSUMSI ENERGI LISTRIK DI GEDUNG JANTO BUILDING

Abstract

Electrical energy consumption in commercial buildings has become a strategic issue in sustainable energy resource management, with Indonesia's building sector contributing 40% of total national energy consumption. This research develops an energy consumption forecasting system based on ARIMA model integrated with machine learning for Janto Building, addressing limitations of manual systems prone to human error and prediction inaccuracy. The methodology employs Research and Development (R&D) approach with ARIMA algorithm implementation, encompassing data preprocessing, optimal parameter identification (p,d,q), and performance evaluation using MSE, MAE, and MAPE metrics. The system was developed using Python with Django framework and SQLite3 database, featuring a user-friendly web interface for parameter configuration and forecasting visualization. Time series forecasting implements a hybrid approach integrating classical statistical techniques and machine learning algorithms for enhanced prediction accuracy. Functional validation through black box and white box testing demonstrates optimal system performance with comprehensive path coverage. Results show the system's capability in generating accurate energy consumption predictions with process automation, reducing manual intervention dependence and supporting national energy reduction targets.