Dynamic Kubernetes Deployment Scaling with Prometheus Metrics

Design a Python script that automatically scales Kubernetes deployments based on custom Prometheus metrics. The solution should dynamically adjust replica counts using a custom algorithm that considers both resource utilization and application-specific performance indicators. Implement advanced logic to prevent oscillation, include exponential backoff strategies, and generate comprehensive logging for each scaling decision. The script must be compatible with standard Kubernetes clusters and integrate seamlessly with existing monitoring infrastructure.