Analisis Penerapan Artificial Neural Network (ANN) pada Indoor Positioning System Berbasis RSSI untuk Lokalisasi Perangkat Kerja

Abstract

Manual management and auditing of work devices often lead to inventory mismatches and time-consuming physical searches. This study analyzes the application of an Artificial Neural Network (ANN) in a Bluetooth Low Energy (BLE)-based Indoor Positioning System (IPS) using Received Signal Strength Indicator (RSSI) as the main input. The model was developed using a Multilayer Perceptron (MLP) architecture with a 3-14-7-3 topology, consisting of 3 input neurons, 14 neurons in the first hidden layer, 7 neurons in the second hidden layer, and 3 output neurons. ReLU was applied in the hidden layers, Softmax in the output layer, and Adam Optimizer for parameter updates. A balanced dataset of 600 samples from three rooms was used with an 80% training and 20% testing split. The testing results show that the ANN model achieved 90.83% accuracy, 90.88% precision, 90.83% recall, 90.85% F1-Score, and 0.1083 MAE. These findings indicate that ANN can effectively learn fluctuating and non-linear RSSI patterns in indoor environments.