A Comprehensive Multi-Label Multi-Instance Learning Prediction Model Using Advanced Computational Intelligence with a Deep Learning Approach

Abstract

Text classification is one of the most prevalent supervised learning tasks. New document classification depends on the availability of a particular training corpus of previously labeled documents. Multi-label text classification is the process of classifying text documents into more than one label concurrently. Machine Learning (ML) models can perform automatic text classification, and Deep Learning (DL) methods are the most suitable tool for multi-label classification of text documents. This study presents a Multi-Instance Learning Prediction using Advanced Deep Learning Model (MILP-ADLM) methodology for document classification. The primary goal was to design a multi-label document classification and sentence labeling system that assigns multiple relevant categories to a document while accurately labeling sentences for their contextual significance. Following that objective, word embedding was performed using Word2Vec, GloVe, and FastText to capture semantic relationships between words. Besides, a Convolutional Transformer Network (CTN) was utilized for classifying documents and labeling sentences, effectively capturing contextual features to enhance the performance of the model. Finally, the Adam optimizer is employed to tune hyperparameters. A comprehensive simulation analysis was performed to determine the performance of the proposed MILP-ADLM model, and the comparative results demonstrated its improved characteristics.