Identification effect of least square fitting method in archives management.

Area of Science:

• Information Science
• Computer Science
• Machine Learning

Background:

• Effective archives management is crucial in the digital information age.
• Accurate identification and classification of archives are essential for system development.
• Existing methods may face challenges in handling complex, non-linear data patterns.

Purpose of the Study:

• To develop an advanced classifier for improved archives identification and classification.
• To enhance the capabilities of Support Vector Machines (SVM) for archival data.
• To introduce a novel approach combining fuzzy logic and LS-SVM.

Main Methods:

• The study utilizes the Least Squares Support Vector Machine (LS-SVM) algorithm.
• A novel wavelet function is incorporated to enhance classifier performance.
• Kernel parameters are optimized using cross-validation techniques.
• Fuzzy theory is integrated with LS-SVM to create the Fuzzy Least Squares Support Vector Machine (FLS-SVM).

Main Results:

• The FLS-SVM classifier demonstrated a classification accuracy of 98.7% on archive datasets.
• The proposed method achieved a low loss rate of only 0.26%.
• Using a wavelet function as the kernel resulted in an average classifier accuracy of 98.38%.

Conclusions:

• The FLS-SVM method proves effective and feasible for archives management, identification, and classification.
• The integration of fuzzy theory and LS-SVM enhances the classification of non-separable data.
• The research validates the utility of least squares fitting methods in archival data processing.