Rule base reduction using conflicting and reinforcement measures