Quantum-inspired attribute selection algorithm: A Fidelity-based Quantum Decision Tree

A classical decision tree is completely based on splitting measures, which utilize the occurrence of random events in correspondence to its class labels in order to optimally segregate datasets. However, the splitting measures are based on greedy strategy, which leads to construction of an imbalanced tree and hence decreases the prediction accuracy of the classical decision tree algorithm. An intriguing approach is to utilize the foundational aspects of quantum computing for enhancing decision tree algorithm. Therefore, in this work, we propose to use fidelity as a quantum splitting criterion to construct an efficient and balanced quantum decision tree. For this, we construct a quantum state using the occurrence of random events in a feature and its corresponding class. The quantum state is further utilized to compute fidelity for determining the splitting attribute among all features. Using numerical analysis, our results clearly demonstrate that the proposed algorithm cooperatively ensures the construction of a balanced tree. We further compared the efficiency of our proposed quantum splitting criterion to different classical splitting criteria on balanced and imbalanced datasets. Our simulation results show that the proposed splitting criterion exceeds all classical splitting criteria for all possible evaluation metrics.