Predicting Corporate Financial Performance Using an Adaptive Neuro-Fuzzy Inference System Algorithm
The present study aimed to predict financial performance using an Adaptive Neuro-Fuzzy Inference System (ANFIS) algorithm. The study was conducted using an exploratory mixed-methods approach (qualitative–quantitative). In the qualitative phase, the fuzzy Delphi technique was employed with the participation of 10 academic and executive experts in the capital market to identify, prioritize, and screen variables affecting financial performance from a large pool of potential indicators. In the quantitative phase, data from 160 companies listed on the Tehran Stock Exchange over the 2016–2024 period (1,440 firm-year observations) were extracted and analyzed using ANFIS to predict “stock price return.” The validity of the final model was assessed through two approaches: quantitatively, by evaluating its generalization accuracy using 421 test observations, and qualitatively, through expert judgment using a validation questionnaire. The fuzzy Delphi results led to the identification of 19 key variables in four categories: financial (7 indicators), economic (6 indicators), social (3 indicators), and macro-environmental (3 indicators). Furthermore, evaluation of the ANFIS model indicated that it explained 75.8% of the variance in stock price return. A comparison of actual and predicted values for 15 selected companies also confirmed that the model successfully tracked the overall trend in returns, although prediction errors increased slightly at extreme values. Accordingly, ANFIS demonstrated satisfactory and practically useful accuracy in predicting financial performance.