优化NetLogo中从CSV加载ABM权重的高效方法问询

问题背景 #

我正在NetLogo中构建基于智能体的模型（ABM），导入了郊区中心、CBD距离、商场、道路等大量GIS数据，用于计算四类智能体（dp=贫困群体、li=低收入群体、mi=中等收入群体、hi=高收入群体）的效用。之前是手动输入权重，现在希望通过可在Excel编辑的外部CSV文件开展敏感性测试，为所有群体配置权重。参考资料编写了加载权重的代码，但受Python编程习惯影响，觉得当前代码效率低下，寻求优化方案。

原NetLogo代码 #

全局变量与基础过程 #

extensions [ csv ]
globals [ 
  data
  dp-weight li-weight mi-weight hi-weight 
  dpu-suburb dpu-neighbour dpu-bu dpu-cbd dpu-mall dpu-markets dpu-road dpu-density dpu-schools dpu-attractive dpu-health dpu-random dpu-water 
  liu-suburb liu-neighbour liu-bu liu-cbd liu-mall liu-markets liu-road liu-density liu-schools liu-attractive liu-health liu-random liu-water
  miu-suburb miu-neighbour miu-bu miu-cbd miu-mall miu-markets miu-road miu-density miu-schools miu-attractive miu-health miu-random miu-water
  hiu-suburb hiu-neighbour hiu-bu hiu-cbd hiu-mall hiu-markets hiu-road hiu-density hiu-schools hiu-attractive hiu-health hiu-random hiu-water
  dp-w-list li-w-list mi-w-list hi-w-list
]
to setup
  clear-all
  file-close-all                          ;; 关闭上次运行后未关闭的文件
  file-open "../data/interim/weights.csv" ;; 打开带表头的权重CSV文件，权重为整数
  set [ dp-w-list li-w-list mi-w-list hi-w-list ]
      [ []        []        []        []        ]
  reset-ticks
end
to go
  load-weights 
  show hiu-water
  tick
end

权重加载过程 #

to load-weights             ;; 基于NetLogo官方CSV示例代码编写
  file-close-all            ;; 关闭所有打开的文件 
  file-open "../data/interim/weights.csv"
  set data csv:from-file "../data/interim/weights.csv"
  set dp-weight item 1 data ;; 跳过首行表头
  set li-weight item 2 data 
  set mi-weight item 3 data 
  set hi-weight item 4 data 
  let i 0
  repeat length dp-weight [
    let w item i dp-weight
    if i > 0 [set dp-w-list lput w dp-w-list] ;; 跳过首列的智能体标识（如'dp'）
    set i i + 1
  ]
  set i 0
  repeat length li-weight [
    let w item i li-weight
    if i > 0 [set li-w-list lput w li-w-list]
    set i i + 1
  ]
  set i 0
  repeat length mi-weight [
    let w item i mi-weight
    if i > 0 [set mi-w-list lput w mi-w-list]
    set i i + 1
  ]
  set i 0
  repeat length hi-weight [
    let w item i hi-weight
    if i > 0 [set hi-w-list lput w hi-w-list]
    set i i + 1
  ]
  set [ dpu-suburb dpu-neighbour dpu-bu dpu-cbd dpu-mall dpu-markets dpu-road dpu-density dpu-schools dpu-attractive dpu-health dpu-random dpu-water ]
      dp-w-list
  set [ liu-suburb liu-neighbour liu-bu liu-cbd liu-mall liu-markets liu-road liu-density liu-schools liu-attractive liu-health liu-random liu-water ]
      li-w-list
  set [ miu-suburb miu-neighbour miu-bu miu-cbd miu-mall miu-markets miu-road miu-density miu-schools miu-attractive miu-health miu-random miu-water ]
      mi-w-list
  set [ hiu-suburb hiu-neighbour hiu-bu hiu-cbd hiu-mall hiu-markets hiu-road hiu-density hiu-schools hiu-attractive hiu-health hiu-random hiu-water ]
      hi-w-list
  file-close                ;; 关闭文件 
end

CSV格式示例 #

agent,suburban,neighbour,bu,cbd,mall,markets,road,density,schools,attractive,health,random,water
dp,3,9,9,4,2,7,3,9,4,2,2,1,10
li,8,7,9,7,2,7,8,9,4,2,4,1,7
mi,6,7,7,4,7,7,6,6,4,7,4,1,5
hi,5,7,2,7,9,6,5,1,4,10,3,1,8

优化方案 #

1. 简化列表提取逻辑 #

原代码中用repeat循环逐个提取权重的写法冗余，直接用but-first跳过首列即可，无需循环：

;; 替换原循环部分
set dp-w-list but-first dp-weight
set li-w-list but-first li-weight
set mi-w-list but-first mi-weight
set hi-w-list but-first hi-weight

一行代码就能替代原来的整个循环块，大幅减少重复代码。

2. 移除冗余的文件操作 #

csv:from-file会自动处理文件打开/关闭，无需手动调用file-open和file-close，原代码中这部分操作属于冗余，直接删除即可。

3. 用表格映射替代全局变量硬编码 #

如果后续需要调整效用维度（比如新增/删除GIS指标），当前大量单独的全局变量会难以维护。可以改用全局变量列表+智能体属性映射的方式：

• 首先定义效用维度的名称列表：

  globals [ weight-dimensions agent-weights ]
  ;; 在setup中初始化维度列表
  set weight-dimensions ["suburban" "neighbour" "bu" "cbd" "mall" "markets" "road" "density" "schools" "attractive" "health" "random" "water"]
  

• 将权重存储为agent-type -> 权重列表的关联表（需启用table扩展）：

  extensions [ csv table ] ;; 新增table扩展
  ;; 在load-weights中构建映射表
  set agent-weights table:make
  foreach but-first data [ ;; 跳过表头行
    let agent-type first ?
    let weights but-first ?
    table:put agent-weights agent-type weights
  ]
  

• 计算效用时直接通过智能体类型查表：

  ;; 假设智能体有agent-type属性，计算效用示例
  to-report calculate-utility [ agent-type ]
    let weights table:get agent-weights agent-type
    ;; 结合GIS数据计算，示例：
    let utility (item 0 weights) * suburban-value + (item 1 weights) * neighbour-value + ...
    report utility
  end
  

这种方式无需定义数十个单独的全局变量，扩展性更强，新增维度或智能体类型时只需修改CSV和维度列表即可。

4. 避免重复加载权重 #

原go过程每次都会调用load-weights，如果是敏感性测试需要多次加载不同权重文件才合理；如果只是初始化加载，应该把load-weights放到setup中，避免每次tick重复读取文件。

优化后的完整load-weights代码 #

to load-weights
  set data csv:from-file "../data/interim/weights.csv"
  ;; 简化列表提取
  set dp-w-list but-first item 1 data
  set li-w-list but-first item 2 data
  set mi-w-list but-first item 3 data
  set hi-w-list but-first item 4 data
  ;; 变量赋值部分保留（如果仍需要单独全局变量）
  set [ dpu-suburb dpu-neighbour dpu-bu dpu-cbd dpu-mall dpu-markets dpu-road dpu-density dpu-schools dpu-attractive dpu-health dpu-random dpu-water ] dp-w-list
  set [ liu-suburb liu-neighbour liu-bu liu-cbd liu-mall liu-markets liu-road liu-density liu-schools liu-attractive liu-health liu-random liu-water ] li-w-list
  set [ miu-suburb miu-neighbour miu-bu miu-cbd miu-mall miu-markets miu-road miu-density miu-schools miu-attractive miu-health miu-random miu-water ] mi-w-list
  set [ hiu-suburb hiu-neighbour hiu-bu hiu-cbd hiu-mall hiu-markets hiu-road hiu-density hiu-schools hiu-attractive hiu-health hiu-random hiu-water ] hi-w-list
end

内容的提问来源于stack exchange，提问作者Alexandre Pereira