Machine learning and bifurcation analysis in a discrete predator-prey model with neem-induced mortality

Area of Science:

• Ecology
• Mathematical Biology
• Computational Science

Background:

• Guava production faces significant pest challenges, necessitating effective and sustainable control strategies.
• Predator-prey models are valuable tools for understanding and managing pest dynamics in agricultural systems.

Purpose of the Study:

• To develop and analyze a discrete-time predator-prey model for guava pest management.
• To investigate the impact of neem-induced mortality and intervention frequency on system stability.
• To explore the application of machine learning for approximating model stability regions.

Main Methods:

• A discrete-time predator-prey model incorporating logistic prey growth, neem effects, and predator interference was formulated using the piecewise constant argument (PCA) scheme.
• Analytical techniques (bifurcation analysis, Lyapunov exponents) and numerical simulations (bifurcation diagrams) were employed to determine stability conditions.
• Machine learning classifiers (Random Forest, Decision Tree) were utilized to approximate the analytically derived stability regions.

Main Results:

• The model identified conditions for flip and Neimark-Sacker bifurcations, revealing complex population dynamics.
• Ecologically, low neem-induced mortality destabilizes coexistence, while higher levels restore stability; intervention frequency critically influences stability, with moderate use promoting it.
• Machine learning models accurately replicated the stability map, with Random Forest showing superior performance, demonstrating ML's potential as a computational surrogate.

Conclusions:

• Sustainable guava pest control hinges on balanced neem application, strategic intervention timing, and the conservation of natural predators.
• The study highlights the interplay between ecological factors and control strategies, providing insights for integrated pest management.
• Machine learning offers a promising avenue for efficiently analyzing complex ecological models and guiding pest management decisions.