Paper: arXiv 2512.17936

Authors: Safiye Turgay, Serkan Erdoğan, Željko Stević, Orhan Emre Elma, Tevfik Eren, Zhiyuan Wang, Mahmut Baydaş

Abstract

In the face of increasing financial uncertainty and market complexity, this study presents a novel risk-aware financial forecasting framework that integrates advanced machine learning techniques with intuitionistic fuzzy multi-criteria decision-making (MCDM). Tailored to the BIST 100 index and validated through a case study of a major defense company in Türkiye, the framework fuses structured financial data, unstructured text data, and macroeconomic indicators to enhance predictive accuracy and robustness. It incorporates a hybrid suite of models, including extreme gradient boosting (XGBoost), long short-term memory (LSTM) network, graph neural network (GNN), to deliver probabilistic forecasts with quantified uncertainty. The empirical results demonstrate high forecasting accuracy, with a net profit mean absolute percentage error (MAPE) of 3.03% and narrow 95% confidence intervals for key financial indicators. The risk-aware analysis indicates a favorable risk-return profile, with a Sharpe ratio of 1.25 and a higher Sortino ratio of 1.80, suggesting relatively low downside volatility and robust performance under market fluctuations. Sensitivity analysis shows that the key financial indicator predictions are highly sensitive to variations of inflation, interest rates, sentiment, and exchange rates. Additionally, using an intuitionistic fuzzy MCDM approach, combining entropy weighting, evaluation based on distance from the average solution (EDAS), and the measurement of alternatives and ranking according to compromise solution (MARCOS) methods, the tabular data learning network (TabNet) outperforms the other models and is identified as the most suitable candidate for deployment. Overall, the findings of this work highlight the importance of integrating advanced machine learning, risk quantification, and fuzzy MCDM methodologies in financial forecasting, particularly in emerging markets.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced machine learning (LSTM, GNN, BNN) and sophisticated fuzzy multi-criteria decision-making (MCDM) methods, indicating high mathematical complexity. It is backed by a detailed empirical case study on BIST 100 with specific metrics (Sharpe ratio, MAPE), demonstrating strong data and implementation focus.

Research Flowchart

  flowchart TD
  A["Research Goal: Develop Risk-Aware Financial Forecasting Framework for Emerging Markets"] --> B["Data Integration & Feature Engineering"]
  B --> C["Computational Modeling: XGBoost, LSTM, GNN, TabNet"]
  C --> D{"Probabilistic Forecasting & Uncertainty Quantification"}
  D --> E{"Intuitionistic Fuzzy MCDM"}
  E --> F["Key Outcomes: High Accuracy MAPE 3.03%, Favorable Risk-Return Sharpe 1.25, TabNet Selected for Deployment"]
  B --> B1["Structured Financial Data"]
  B --> B2["Unstructured Text Data"]
  B --> B3["Macroeconomic Indicators"]
  E --> E1["Entropy Weighting"]
  E --> E2["EDAS & MARCOS Ranking"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a risk-aware AI-enhanced financial forecasting framework that integrates advanced machine learning models and intuitionistic fuzzy MCDM for the Turkish BIST 100 market.

Source sentence
In this paper, we propose a risk-aware, AI-enhanced financial forecasting framework tailored to the Turkish BIST 100 index.

Assumes 3

  • Robust financial forecasting must incorporate nonlinear relationships, exogenous risk factors, and predictive uncertainty. model
    Source sentence
    A basic hypothesis of the framework is that any robust financial forecasting needs to incorporate nonlinear relationships, exogenous risk factors, and predictive uncertainty.
  • Forecast errors follow a normal distribution providing a full probability distribution for the target variable. statistical
    Source sentence
    A risk-aware forecast provides a full probability distribution for the target variable. More formally, the forecast is given by
  • Missing financial data (less than 5% of entries) can be accurately imputed using KNN with k=5. data
    Source sentence
    Missing financial data, less than 5% of all entries, was imputed using the KNN approach with k = 5, since it is superior to other mean-based or median-based approaches in terms of maintaining multivariate relationships.

Data

Backtest sample: BIST 100 · equities

  • Company financial reports (quarterly and annual balance sheets) · Public Disclosure Platform (KAP) fundamentals public
    2018-2024 · BIST 100 firms
    Source sentence
    structured data is obtained by collecting company financial reports, such as quarterly and annual balance sheets from the BIST 100 firms, through the Public Disclosure Platform (KAP, kap.org.tr/en, for the Turkish capital markets).
  • Macroeconomic indicators (inflation, interest rates, exchange rates, GDP growth) · Turkish Statistical Institute (TÜİK) and Central Bank of the macro public
    2018-2024
    Source sentence
    macroeconomic indicators-including inflation, interest rates, exchange rates, and gross domestic product (GDP) growth obtained from the Turkish Statistical Institute (TÜİK) and the Central Bank of the Republic of Türkiye (TCMB).
  • Financial news articles and earnings-call transcripts · KAP and major national news agencies text-news public
    2018-2024 · Turkish defense sector
    Source sentence
    more than 15,000 financial news articles and 120 earnings-call transcripts related to the Turkish defense sector have been gathered from KAP and major national news agencies.

Costs

Not applicable The paper reports no trading or execution result.

Finds 5

Reported performance (Hybrid risk-aware financial forecasting framework (TabNet)): Sharpe 1.25

  • XGBoost predicts net profit with a MAPE of 3.03%, while TabNet reduces this error to 2.91%, both outperforming the ARIMA benchmark. predictive-accuracy
    Source sentence
    The core XGBoost implementation achieved a net profit MAPE of 3.03%, while the best-performing model (i.e., TabNet) within the extended framework reduced this to 2.91%.
  • The model yields narrow 95% confidence intervals for total assets, net profit, and equity, indicating low forecast uncertainty. statistical
    Source sentence
    The 95% confidence intervals of total assets, net profit, and equity are relatively narrow, indicating that the model’s forecasts are subject to a low degree of uncertainty under current market conditions.
  • Risk‑adjusted performance metrics are favorable: Sharpe Ratio = 1.25 and Sortino Ratio = 1.80. performance
    Source sentence
    The Sharpe Ratio calculated equals 1.25, whereas the calculated value of the Sortino Ratio is 1.80.
  • MAE for total assets is 50 million TRY and MAPE for net profit is 3.03%, demonstrating good forecasting accuracy. predictive-accuracy
    Source sentence
    The MAPE of net profit is 3.03%, which means a good level of forecasting accuracy.
  • Sensitivity analysis shows key financial indicators are highly responsive to inflation, interest rates, sentiment, and exchange rates. other
    Source sentence
    Sensitivity analysis also showed that key forecasted financial indicators responded sensitively to changes in inflation, interest rates, sentiment, and exchange rates.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 65 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

No repository by the authors found in the paper.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.