In today's economy, competition has reached a global scale. Consequently, the management of working capital is critical for businesses to maintain uninterrupted operations. Due to its significance in corporate finance specifically regarding its share of total assets, its role in supporting sales, and its impact on profitability the effective management of working capital is a vital area of study. This research examines the performance metrics of Return on Assets (RoA) a widely used financial metric in corporate finance literature using deep learning and machine learning methods. Data covering the 2015–2025 quarterly periods for 127 manufacturing firms listed on the Borsa Istanbul (BIST) were utilized; the dataset included usage indices, performance indices, efficiency indices, cash conversion cycles, leverage ratios, and RoA figures. Analysis results indicate that the error in predicting ROA for the subsequent quarter decreased by 31.8% compared to the naïve model. For one-year-ahead predictions, the improvement was 12.5%. The indices may offer short-term incremental returns. In one-year forecasts, some models showed limited predictive contribution. The study presents comparable results obtained from deep learning and tree-based models.
Return on assest, working capital management efficiency, deep learning, machine learning