This study proposes a hybrid Data Envelopment Analysis-Backpropagation Neural Network (DEA-BPNN) framework to evaluate and predict the efficiency performance of furniture and wood processing firms listed on the Indonesia Stock Exchange (IDX). As a strategic manufacturing sector, firm performance in this industry is frequently challenged by cost volatility, scale inefficiencies, and fluctuating market demand, while conventional efficiency methods remain limited in capturing nonlinear relationships and predictive insights. To address these limitations, the study integrates frontier-based efficiency measurement with machine learning-based prediction. Using panel data from six IDX-listed firms over the 2020-2024 period, efficiency scores are first estimated through CCR and BCC DEA models, with total assets, cost of goods sold, and operating expenses as inputs, and revenue and profit as outputs. The results reveal notable heterogeneity in efficiency performance, where several firms achieve full BCC efficiency, indicating strong pure technical efficiency, while variations in CCR efficiency highlight the presence of scale inefficiencies. In the second stage, a BPNN model is developed to predict CCR and BCC efficiency scores. The optimized 5-8-2 network architecture demonstrates strong predictive performance, achieving a Mean Squared Error (MSE) of 0.0145, low Mean Absolute Percentage Error (MAPE) values of 1.22% (CCR) and 0.89% (BCC), and high Pearson correlation coefficients of 0.94 and 0.96. Overall, the findings confirm that the hybrid DEA-BPNN framework provides a robust tool for efficiency evaluation and prediction, supporting performance monitoring and strategic decision-making in Indonesia’s furniture and wood processing industry.