name: agent-based-simulator description: 基于智能体的建模技能，用于模拟具有异质交互智能体的复杂适应系统 allowed-tools:

Read
Write
Glob
Grep
Bash metadata: specialization: 决策智能 domain: 商业 category: 仿真 priority: lower tools-libraries:
- mesa
- agentpy
- pyNetLogo

基于智能体的模拟器

概述

基于智能体的模拟器技能提供通过模拟异质、交互的智能体来建模复杂适应系统的能力。它支持自下而上地理解涌现的市场行为、客户动态和竞争互动，为战略决策提供支持。

能力

智能体定义与行为建模
环境与空间建模
交互规则设定
涌现行为观察
参数扫描
集成模拟运行
可视化与动画
结果统计分析

使用流程

战争推演与竞争响应建模
市场规模与机会评估
客户细分分析

使用方法

智能体定义

# 定义客户智能体
customer_agent = {
    "type": "Customer",
    "attributes": {
        "budget": {"distribution": "normal", "mean": 1000, "std": 200},
        "brand_loyalty": {"distribution": "uniform", "min": 0, "max": 1},
        "price_sensitivity": {"distribution": "beta", "alpha": 2, "beta": 5},
        "preferred_features": ["功能列表"]
    },
    "behaviors": {
        "purchase_decision": {
            "triggers": ["需求产生", "看到促销"],
            "evaluation": "加权效用",
            "factors": ["价格", "质量", "品牌匹配度"]
        },
        "word_of_mouth": {
            "probability": 0.3,
            "reach": {"distribution": "poisson", "lambda": 5},
            "sentiment_spread": True
        },
        "brand_switching": {
            "threshold": 0.7,
            "factors": ["满意度", "竞争对手促销"]
        }
    }
}

环境定义

# 定义市场环境
environment = {
    "topology": "network",  # 或 "grid", "continuous"
    "network_type": "small_world",
    "network_params": {"k": 6, "p": 0.1},
    "global_properties": {
        "economic_condition": {"initial": "normal", "transitions": "markov"},
        "market_size": 10000,
        "growth_rate": 0.02
    }
}

交互规则

# 定义交互规则
interactions = {
    "customer_customer": {
        "information_sharing": {
            "probability": "基于关系",
            "content": ["产品体验", "价格信息"]
        },
        "social_influence": {
            "mechanism": "阈值模型",
            "threshold_distribution": "normal"
        }
    },
    "customer_company": {
        "purchase": {
            "frequency": "基于需求",
            "channel": ["线上", "线下", "混合"]
        },
        "complaint": {
            "trigger": "满意度 < 0.3",
            "resolution_impact": 0.5
        }
    },
    "company_company": {
        "price_competition": "古诺|伯特兰|斯塔克尔伯格",
        "market_signaling": True
    }
}

模拟配置

# 模拟设置
simulation_config = {
    "time_steps": 365,
    "agents": {
        "Customer": 5000,
        "Company": 3
    },
    "ensemble_runs": 100,
    "parameter_sweep": {
        "price_sensitivity_mean": [0.3, 0.5, 0.7],
        "word_of_mouth_probability": [0.1, 0.3, 0.5]
    },
    "data_collection": {
        "agent_level": ["satisfaction", "brand_choice"],
        "model_level": ["market_shares", "total_revenue", "gini_coefficient"]
    }
}

输入模式

{
  "agents": {
    "agent_type": {
      "count": "number",
      "attributes": "object",
      "behaviors": "object"
    }
  },
  "environment": {
    "topology": "string",
    "properties": "object"
  },
  "interactions": "object",
  "simulation_config": {
    "time_steps": "number",
    "ensemble_runs": "number",
    "parameter_sweep": "object",
    "random_seed": "number"
  }
}

输出模式

{
  "summary_statistics": {
    "metric_name": {
      "mean": "number",
      "std": "number",
      "percentiles": "object",
      "time_series": ["number"]
    }
  },
  "emergent_patterns": [
    {
      "pattern": "string",
      "frequency": "number",
      "conditions": "object"
    }
  ],
  "parameter_sweep_results": {
    "parameter_combination": {
      "outcomes": "object"
    }
  },
  "agent_trajectories": "object (sample)",
  "network_metrics": {
    "clustering_coefficient": "number",
    "average_path_length": "number",
    "degree_distribution": "object"
  },
  "visualization_paths": ["string"]
}

最佳实践

从简单的智能体开始，逐步增加复杂性
根据现实世界观察验证智能体行为
使用集成运行来考虑随机变异性
对关键参数进行敏感性分析
记录所有行为规则及其合理性
在极端条件下测试涌现行为
在可用时与聚合级数据进行比较

使用案例

使用案例	智能体类型	关键行为
市场动态	客户、公司	购买、定价
创新扩散	采纳者、影响者	采纳、沟通
供应链	供应商、分销商、零售商	订购、库存
意见形成	公民、媒体	影响、信息传播

集成点

与系统动力学建模器连接，用于混合方法
为竞争场景提供战争推演编排器
支持场景叙事生成器，用于故事线创建
与蒙特卡洛引擎集成，用于不确定性传播