نوع مقاله : مقاله پژوهشی
نویسندگان
1 شهرداری تهران ، اداره کل خزانه داری
2 هیات علمی دانشگاه
3 عضو هیات علمی
4 Azad University, Islamshahr Branch
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Objective
The present study aims to propose a framework for explainable data mining to support behavioral data-driven decision-making in investor portfolio selection. This research seeks to go beyond mere prediction and transform behavioral data into understandable and actionable information for decision-makers.
Methodology
This study utilized data from 400 active capital market investors and analyzed their psychological and behavioral characteristics using machine learning algorithms. To compare model performance, five-fold cross-validation was employed, and models were evaluated using Accuracy, Precision, Recall, F1-score, and Area Under the Curve (AUC). To explain model outputs, feature importance analysis and model explainability analysis based on SHAP (SHapley Additive exPlanations) values were applied.
Findings
The results showed that the XGBoost model achieved the best performance among the examined models, with an Accuracy of 0.86 and an AUC of 0.81. Furthermore, Risk Perception with a relative importance of 0.22, Information Processing with 0.19, and Responsibility Attribution with 0.17 were identified as the most important features of the model. The explainability analysis revealed that the decision-making structure of investors differs between those with internal and external loci of control. Specifically, in the group with an internal locus of control, analytical information processing, responsibility attribution, and learning from past outcomes were more prominent, whereas in the group with an external locus of control, risk perception and reaction to market uncertainty played a greater role. Additionally, SHAP analysis uncovered the existence of individual differences within the groups.
Conclusion
The findings demonstrate that the combination of behavioral data, machine learning, and the SHAP method can go beyond mere prediction and transform behavioral data into explainable and actionable information for decision support and portfolio selection. This approach can serve as a foundation for developing personalized decision-support systems in the investment domain
Keywords: Explainable Data Mining, Machine Learning, SHAP, Behavioral Data, Investment Decision-Making, Investor Behavior, Portfolio Selection, Locus of Control.
کلیدواژهها [English]
)
ارسال نظر در مورد این مقاله