Computational Intelligence-Based Financial Crisis Prediction Model Using Feature Subset Selection with Optimal Deep...

Area of Science:

• Computational intelligence
• Machine learning
• Financial modeling

Background:

• Financial crisis prediction (FCP) relies on classifier techniques to categorize institutional data.
• Effective FCP requires careful feature selection (FS) to improve classifier performance.
• Computational intelligence offers robust classification models for financial risk assessment.

Purpose of the Study:

• To propose a novel feature selection (FS) and classification approach for financial crisis prediction (FCP).
• To enhance FCP accuracy by integrating elephant herd optimization (EHO) for FS and modified water wave optimization (MWWO)-deep belief network (DBN) for classification.

Main Methods:

• Feature selection (FS) performed using the elephant herd optimization (EHO) algorithm.
• Classification model based on a deep belief network (DBN) optimized with the modified water wave optimization (MWWO) algorithm.
• Proposed MWWO-DBN model evaluated using benchmark datasets: AnalcatData, German Credit, and Australian Credit.

Main Results:

• The EHO algorithm effectively identified optimal feature subsets.
• The MWWO algorithm successfully tuned deep belief network (DBN) parameters.
• The integrated EHO-MWWO-DBN model demonstrated superior classification performance on benchmark datasets.

Conclusions:

• The proposed EHO-based FS combined with MWWO-DBN classification significantly improves financial crisis prediction.
• This hybrid approach offers a promising solution for accurate institutional financial risk assessment.
• The method highlights the potential of advanced computational intelligence techniques in financial forecasting.