AN EFFICIENT METHOD FOR MULTICLASS CLASSIFICATION IN TABULAR DATA USING MULTILAYER PERCEPTRON MODEL

Abstract

In recent years, neural networks have shown remarkable promise in addressing various classification challenges, especially in fields like image and text processing. However, their effectiveness on structured tabular data remains relatively underexplored. This research examines the use of Multilayer Perceptron (MLP) neural networks for multiclass classification tasks on tabular datasets. Preprocessing steps, including feature scaling and one-hot encoding, were applied to enhance performance. Different MLP architectures were evaluated, with variations in the number of hidden layers and neurons, as well as the use of regularization methods such as dropout to mitigate overfitting. Early stopping was also utilized to improve generalization. The model was trained using categorical cross-entropy loss and the Adam optimizer, achieving a good balance between training accuracy and validation results. Findings suggest that MLPs can be effectively applied to multiclass classification on tabular data, though performance is highly dependent on hyperparameter selection and preprocessing techniques. This study outlines key approaches for optimizing MLP architectures for structured data and provides insights into best practices for neural network classification tasks in similar contexts.