Research Projects

Inferring causal relationships in the decision-making processes of machine learning algorithms is a crucial step toward achieving explainable Artificial Intelligence (AI). In this research proposal, we investigate the development of a novel causality measure derived from Lempel-Ziv (LZ) complexity. We explore how the proposed causality measure can be used in decision trees by enabling splits based on features that most strongly cause the outcome. We further evaluate the effectiveness of the causality-based decision tree in comparison to a traditional decision tree using Gini impurity. Based on the features used in the LZ causal measure based decision tree, we plan to introduce a causal strength for each feature in the dataset so as to infer the predominant causal variables for the occurrence of the outcome. The main objectives of the research proposal are highlighted : Develop a Novel Causal Measure: Create a new model agnostic causal measure using compression complexity measures such as Lempel-Ziv Complexity to infer causal direction from univariate temporal and non-temporal datasets. Test and Validate the Measure: Evaluate the proposed measure's effectiveness on both synthetic and real-world datasets. Integrate Causality in Decision Trees: Incorporate the causal measure into decision tree models as a splitting criterion, enabling causal-informed decision-making at each node. Performance Comparison: Benchmark the proposed methods against classical decision trees using the Gini impurity criterion on multiple datasets, including AR, Iris, Breast Cancer, Voting, and others. Feature Importance via Causal Influence: Introduce a feature importance score based on the proposed LZ causal measure, providing an interpretable ranking of features by their causal influence on the outcome. Collaboration with Domain Experts: The interpretability of the proposed decision tree, developed using the novel Lempel-Ziv complexity-based causality measure, will be evaluated in collaboration with domain experts. This effort will focus primarily on healthcare datasets, including those related to liver cirrhosis and breast cancer. By incorporating expert feedback, we aim to ensure that the model’s decisions are both scientifically sound and practically meaningful. The successful completion of this project will pave the way for deploying the proposed tool in clinical settings, enhancing decision-making processes and supporting medical professionals in diagnosis and treatment planning.