Yazar "Zivkovic, Miodrag" seçeneğine göre listele

Listeleniyor 1 - 2 / 2
  • Küçük Resim Yok
    Öğe
    Improving audit opinion prediction accuracy using metaheuristics-tuned XGBoost algorithm with interpretable results through SHAP value analysis
    (Elsevier, 2023) Todorovic, Mihailo; Stanisic, Nemanja; Zivkovic, Miodrag; Bacanin, Nebojsa; Simic, Vladimir; Tirkolaee, Erfan Babaee
    This study aims to create a machine learning model that can predict opinions in external audits and surpass the benchmark set in a prior study from the literature. This tool could reduce audit risk, which is a crucial task in external audits. Previous studies have shown that it is possible to create models that can predict the audit opinion a company will receive. In these studies, authors used statistics and machine learning models, and both non-financial (e.g. audit lag) and financial data (e.g. financial ratios, or absolute value items available from financial statements) to make predictions. In this study, the performance of the XGBoost model optimized by metaheuristics algorithms is examined and evaluated. This study compares the performance of six different metaheuristic algorithms used to tune the XGBoost model in two separate scenarios. The first scenario represents a realistic client portfolio, where a majority of the clients are known, while the second scenario simulates a new clients-only portfolio, a more difficult scenario where prior information such as audit lag is not available. The study uses a dataset of 12,690 observations of Serbian companies and their audit opinions from 2016 to 2019. The findings indicate an improvement over the benchmark due to a more optimized hyperparameter tuning process and the use of the iterative sine-cosine algorithm for the XGBoost model.
  • Küçük Resim Yok
    Öğe
    Optimizing long-short-term memory models via metaheuristics for decomposition aided wind energy generation forecasting
    (Springer, 2024) Pavlov-Kagadejev, Marijana; Jovanovic, Luka; Bacanin, Nebojsa; Deveci, Muhammet; Zivkovic, Miodrag; Tuba, Milan; Strumberger, Ivana
    Power supply from renewable energy is an important part of modern power grids. Robust methods for predicting production are required to balance production and demand to avoid losses. This study proposed an approach that incorporates signal decomposition techniques with Long Short-Term Memory (LSTM) neural networks tuned via a modified metaheuristic algorithm used for wind power generation forecasting. LSTM networks perform notably well when addressing time-series prediction, and further hyperparameter tuning by a modified version of the reptile search algorithm (RSA) can help improve performance. The modified RSA was first evaluated against standard CEC2019 benchmark instances before being applied to the practical challenge. The proposed tuned LSTM model has been tested against two wind production datasets with hourly resolutions. The predictions were executed without and with decomposition for one, two, and three steps ahead. Simulation outcomes have been compared to LSTM networks tuned by other cutting-edge metaheuristics. It was observed that the introduced methodology notably exceed other contenders, as was later confirmed by the statistical analysis. Finally, this study also provides interpretations of the best-performing models on both observed datasets, accompanied by the analysis of the importance and impact each feature has on the predictions.