如何将时序多层多路有向GraphML文件导入NetLogo？

一、预处理GraphML文件（Python脚本） #

NetLogo原生解析带复杂属性的时序GraphML容易出现解析偏差，先用Python提取结构化数据，转成NetLogo易读取的CSV格式：

import networkx as nx
import pandas as pd

# 读取时序GraphML文件
G = nx.read_graphml("your_temporal_graph.graphml")

# 提取静止节点数据：节点ID + 所有属性
nodes_data = []
for node_id, attrs in G.nodes(data=True):
    node_row = {"node_id": node_id}
    node_row.update(attrs)
    nodes_data.append(node_row)
pd.DataFrame(nodes_data).to_csv("static_nodes.csv", index=False)

# 提取边与车辆关联数据：源节点、目标节点、veh_type、edgeID、时间戳等
edges_data = []
for u, v, attrs in G.edges(data=True):
    edge_row = {"source": u, "target": v}
    edge_row.update(attrs)
    edges_data.append(edge_row)
pd.DataFrame(edges_data).to_csv("edges_with_vehicles.csv", index=False)

二、NetLogo导入实现 #

1. 定义核心breed与全局变量 #

; 静止节点海龟breed
breed [static-nodes static-node]
; 移动车辆海龟breed
breed [vehicles vehicle]
; 两种有向链接breed（对应veh_type的两类）
directed-link-breed [car-links car-link]
directed-link-breed [truck-links truck-link]

; 存储节点ID到海龟的映射，用于快速关联边
globals [node-id-map]

2. 导入静止节点 #

to import-static-nodes
  set node-id-map make-dictionary
  file-open "static_nodes.csv"
  ; 跳过CSV表头
  let header file-read-line
  while [not file-at-end?] [
    let line file-read-line
    let parts split line ","
    let node-id item 0 parts
    ; 创建静止节点，若GraphML含坐标属性，可替换随机布局为对应属性值
    create-static-nodes 1 [
      set xcor random-xcor
      set ycor random-ycor
      set label node-id
      dict-put node-id-map node-id self
    ]
  ]
  file-close
end

3. 导入边与对应车辆海龟 #

to import-edges-and-vehicles
  file-open "edges_with_vehicles.csv"
  let header file-read-line
  while [not file-at-end?] [
    let line file-read-line
    let parts split line ","
    let source-id item 0 parts
    let target-id item 1 parts
    let veh-type item 2 parts ; 根据实际CSV列索引调整
    let edge-id item 3 parts ; 根据实际CSV列索引调整
    
    ; 获取源、目标节点的海龟实例
    let source-turtle dict-get node-id-map source-id
    let target-turtle dict-get node-id-map target-id
    
    ; 根据veh_type创建对应有向链接
    if veh-type = "car" [
      create-car-link-from source-turtle to target-turtle [
        set label edge-id
      ]
    ]
    if veh-type = "truck" [
      create-truck-link-from source-turtle to target-turtle [
        set label edge-id
      ]
    ]
    
    ; 创建对应edgeID的车辆海龟，默认放在链接中间位置
    create-vehicles 1 [
      set xcor ( [xcor] of source-turtle + [xcor] of target-turtle ) / 2
      set ycor ( [ycor] of source-turtle + [ycor] of target-turtle ) / 2
      set label edge-id
    ]
  ]
  file-close
end

4. 时序切换逻辑（可选） #

针对时序图需求，可添加时间戳控制逻辑，切换不同时间步的图结构：

globals [current-timestamp]

to setup-temporal-graph
  import-static-nodes
  set current-timestamp 0
end

to go-to-timestamp [target-ts]
  set current-timestamp target-ts
  ; 清除当前时间步的链接与车辆
  ask vehicles [ die ]
  ask car-links [ die ]
  ask truck-links [ die ]
  
  file-open "edges_with_vehicles.csv"
  let header file-read-line
  while [not file-at-end?] [
    let line file-read-line
    let parts split line ","
    let ts item 4 parts ; 假设时间戳在CSV第5列，按需调整
    if ts = word target-ts [
      ; 重复边与车辆的创建逻辑
      let source-id item 0 parts
      let target-id item 1 parts
      let veh-type item 2 parts
      let edge-id item 3 parts
      let source-turtle dict-get node-id-map source-id
      let target-turtle dict-get node-id-map target-id
      
      if veh-type = "car" [
        create-car-link-from source-turtle to target-turtle [ set label edge-id ]
      ]
      if veh-type = "truck" [
        create-truck-link-from source-turtle to target-turtle [ set label edge-id ]
      ]
      
      create-vehicles 1 [
        set xcor ( [xcor] of source-turtle + [xcor] of target-turtle ) / 2
        set ycor ( [ycor] of source-turtle + [ycor] of target-turtle ) / 2
        set label edge-id
      ]
    ]
  ]
  file-close
end

三、关键注意事项 #

• 调整CSV列索引：需根据Python导出的CSV表头，修改NetLogo代码中item的索引值，确保属性匹配。
• 节点位置配置：若GraphML含节点坐标属性，直接替换随机布局代码为对应属性读取逻辑即可。
• 多层图适配：若原GraphML含多层属性，预处理时需提取layer字段，导入时可通过颜色、标签区分不同层的元素。

内容的提问来源于stack exchange，提问作者pandora