Feature selection plays a critical role in optimizing machine learning models by reducing dimensionality, improving classification performance, and enhancing computational efficiency. This study proposes the Adaptive Quantum Flora Feature Selection (AQFFS) algorithm, which combines artificial flora-inspired growth dynamics with quantum probability-based transitions to identify compact and informative feature subsets. Structured refinement is achieved through root, stem, and leaf growth phases, while Hadamard and Pauli-X quantum gates introduce probabilistic transitions that balance exploration and exploitation. AQFFS was evaluated on 16 benchmark datasets using both a Quantum Circuit-based classifier and the K-Nearest Neighbors (KNN) classifier to assess performance across quantum and classical paradigms. Comparative experiments against several state-of-the-art feature selection methods, including Artificial Flora Optimization (AFO), Quantum Genetic Algorithm (QGA), Quantum Particle Swarm Optimization (QPSO), Improved Binary Manta Ray Foraging Optimization (IBMRFO), Fuzzy Fitness Memetic Algorithm with Tabu Search and Hill Climbing (FFMATSHC), and Fuzzy PSO with Greedy Forward Selection (FPGFS), demonstrated improvements in classification accuracy, feature subset reduction, and runtime efficiency. Ablation analysis highlighted the importance of both operator-inspired stochastic updates and biologically guided adaptation, while statistical significance testing confirmed the consistency of performance gains across datasets and classifiers. The results indicate that AQFFS provides a scalable and model-agnostic framework for high-dimensional feature selection, with applicability in both classical and quantum machine learning contexts.