Skip to content
Publications

Talk · IEEE International Conference on Problems of Informatics, Electronics and Radio Engineering · November 15, 2024

Explainable Artificial Intelligence Methods in Text Classification Machine Learning Models for Investigating Human Behavior in Digital Environments

Fail Mubarakovich Gafarov, Viliuza R. Gafarova, Pavel Nikolaevich Ustin

Abstract

Text classification, also commonly known as text categorization, is the process of assigning labels or tags to textual data. To solve this problem different models are used, based on machine learning algorithms, as well as deep learning methods based on neural networks of different architectures (recurrent, convolutional, transformers, etc.). But most complex machine learning and deep learning models are black boxes because they contain a large number of parameters and have complex architectures that make them uninterpretable. This paper proposes a framework to generate explanations in binary text classification task for an investigation of subject's behavior in the digital environment. Our framework utilizes five machine learning models (Random Forest, K-nearest neighbors, Decision Tree, Logistic Regression, Gradient Boosting) and a fine-tuned BERT-based classifier and an explainability module using LIME, SHAP and Integrated Gradients to generate explanations for each prediction. The main objective of this paper is to show how machine learning-based natural language processing methods enhanced by Explainable AI methods can be applied in psychological domain.

Explainable Artificial Intelligence (XAI)Text and Document Classification TechnologiesMental Health via Writingexplainable AItext classificationbinary text classification

More by Fail Gafarov