Application of forward selection strategy using C4.5 algorithm to improve the accuracy of classification's data set.

Area of Science:

• Machine Learning
• Bioinformatics
• Data Mining

Background:

• Accurate breast cancer classification is crucial for effective treatment.
• Traditional machine learning algorithms like C4.5 and Naïve Bayes have limitations in classification accuracy.
• Feature selection is a key technique to improve model performance.

Purpose of the Study:

• To enhance the classification accuracy of the C4.5 algorithm.
• To evaluate the effectiveness of a forward selection technique combined with C4.5 for breast cancer data.
• To compare the performance of the proposed C4.5 + Forward Selection (FS) model against C4.5 and Naïve Bayes.

Main Methods:

• Utilized the Breast Cancer dataset from the UCI Machine Learning Repository.
• Applied a feature selection technique (weight by information gain) followed by forward selection.
• Split the dataset into training and testing sets with ratios of 70:30, 80:20, and 90:10.
• Evaluated models using RapidMiner, comparing accuracy, precision, recall, F-measure, and Area Under Curve (AUC).

Main Results:

• The C4.5 + FS model outperformed standard C4.5 and Naïve Bayes in classification accuracy.
• C4.5 + FS achieved accuracy values of 76.74% (70:30), 78.95% (80:20), and 78.57% (90:10).
• Average accuracy increased by 12.97% over C4.5 and 8.32% over Naïve Bayes.
• C4.5 + FS demonstrated superior precision (79.84%), recall (92.50%), and F-measure (85.55%).
• The average Area Under Curve (AUC) improved from 0.628 to 0.732%.

Conclusions:

• The forward selection strategy effectively enhances the C4.5 algorithm's classification accuracy for the Breast Cancer dataset.
• The proposed C4.5 + FS method offers improved performance metrics, making it a valuable tool for breast cancer analysis.
• Feature selection is a viable approach to optimize machine learning models in medical data classification.