Early Warning of Enterprise Financial Risk Based on Decision Tree Algorithm.

Area of Science:

• Computer Science
• Data Science
• Financial Technology

Background:

• Traditional decision tree algorithms face limitations in enterprise financial early warning.
• Existing Principal Component Analysis (PCA)-based decision tree algorithms show suboptimal data representativeness after dimensionality reduction, limiting accuracy improvements.

Purpose of the Study:

• To propose a novel, improved decision tree algorithm for enhancing enterprise financial early warning systems.
• To address the limitations of existing algorithms by improving data representativeness and classification accuracy.

Main Methods:

• Developed an improved decision tree algorithm by extracting attribute eigenvalues twice before classification.
• Implemented a secondary dimension reduction and merging selection process after subtree establishment.
• Validated the algorithm using three datasets from the UCI database.

Main Results:

• The improved algorithm achieved an average accuracy of 94.6% across three datasets.
• Demonstrated accuracy improvements of 1.6% over the traditional classical algorithm and 0.6% over the ordinary PCA decision tree optimization algorithm.
• Confirmed the practical importance of PCA-based decision tree algorithms in improving prediction accuracy.

Conclusions:

• The proposed decision tree algorithm effectively enhances enterprise financial early warning accuracy.
• The method offers a more robust approach to identifying financial risks compared to existing techniques.
• Future work will focus on secondary modeling for a more efficient decision tree model.