Defining and evaluating classification algorithm for high-dimensional data based on latent topics.

Area of Science:

• Information Retrieval
• Data Mining
• Machine Learning

Background:

• Automatic text categorization is crucial for information retrieval and data mining.
• Existing methods struggle with efficiency on large, high-dimensional datasets.
• There is a need for faster and more effective text classification techniques.

Purpose of the Study:

• To present a novel text categorization method combining Latent Dirichlet Allocation (LDA) and Support Vector Machine (SVM).
• To improve classification efficiency and performance on large datasets.
• To demonstrate a streamlined classification process with broad applicability.

Main Methods:

• Latent Dirichlet Allocation (LDA) was used for dimensionality reduction, extracting semantic topic features.
• Support Vector Machine (SVM) was employed for classification using the reduced feature set.
• The combined LDA+SVM model was evaluated on benchmark datasets (20 Newsgroups, Reuters-21578).

Main Results:

• The LDA+SVM model achieved high performance metrics, including precision, recall, and F1 measure.
• The proposed method significantly reduced classification time compared to existing approaches.
• Experimental results validated the effectiveness and efficiency of the LDA+SVM model.

Conclusions:

• The LDA+SVM model offers a highly efficient and accurate solution for automatic text categorization.
• This approach effectively handles large and high-dimensional datasets, overcoming previous limitations.
• The method shows strong potential for practical applications requiring streamlined text classification.