WeNet-RF: An Automatic Classification Model for Financial Reimbursement Budget Items

Area of Science:

• Financial Management
• Computational Linguistics
• Machine Learning

Background:

• Manual classification of financial reimbursement budget items is inefficient and prone to errors.
• Current methods lack the speed and precision required for regulatory compliance.
• Labor-intensive processes increase the risk of incorrect category selection.

Purpose of the Study:

• To develop an automated model for accurate and efficient classification of financial reimbursement budget items.
• To address the limitations of manual classification in financial management.
• To enhance the legitimacy and regulatory compliance of financial expenditures.

Main Methods:

• Proposed a WeNet-Random Forest (WeNet-RF) model integrating speech recognition (WeNet) and Random Forest (RF).
• The model involves four steps: speech identification, feature extraction, item classification, and model evaluation.
• Compared WeNet-RF performance against Convolutional Neural Networks (CNN), Logistic Regression (LR), and K-Nearest Neighbors (KNN).

Main Results:

• The WeNet-RF model achieved an accuracy rate, precision rate, recall rate, and F1 score of 90.77%.
• Performance was validated using 50 real financial reimbursement records.
• WeNet-RF demonstrated superior efficiency and accuracy compared to other tested models.

Conclusions:

• The WeNet-RF model offers a robust solution for improving financial management processes.
• This automated approach enhances the accuracy and efficiency of budget item classification.
• The study provides a valuable reference model for financial management systems.