Lasso-Based Feature Selection for Enhancing Accuracy and Efficiency in Date Fruit Classification

This study presents an improved method to enhance classification efficiency by combining the Lasso (Least Absolute Shrinkage and Selection Operator) feature selection technique with a Multilayer Perceptron (MLP) model. Lasso selected an optimal subset of 19 from 34 initial features, enabling the MLP model to achieve an impressive accuracy of 96.11%. This result not only surpasses that of complex ensemble learning methods using the full feature set (which previously achieved 95.56% performance) but also demonstrates Lasso’s superiority over other feature selection methods such as Mutual Information (MI) and the ANOVA F-test for this problem. More importantly, this important feature selection not only enhances computational efficiency but also improves predictive accuracy by focusing the model on the most pertinent information, thereby reducing noise and the risk of overfitting. These findings suggest a practical way for developing agricultural quality control solutions that are both highly accurate and computationally efficient.