使用 Networkx 绘制二部图

本文以工厂 - 仓库的运输成本、仓库 - 商品的销量为背景，绘制了能清晰反映结点之间关系的二部图。

数据

工厂 - 仓库的运输成本

factory_id	dc_id	unit_cost
F001	DC001	224
F001	DC002	198
F001	DC003	196
F001	DC004	170
F001	DC005	170
F001	DC006	212
F001	DC007	169
F001	DC008	344
F001	DC009	283
F001	DC010	518
F001	DC011	458
F001	DC012	1452
F001	DC013	499
F001	DC014	438
F001	DC015	443
F001	DC016	396
F001	DC017	260
F001	DC018	248
F002	DC001	444
F002	DC002	434
F002	DC003	356
F002	DC004	428
F002	DC005	418
F002	DC006	399
F002	DC007	439
F002	DC008	477
F002	DC009	463
F002	DC010	390
F002	DC011	397
F002	DC012	875
F002	DC013	343
F002	DC014	268
F002	DC015	361
F002	DC016	378
F002	DC017	282
F002	DC018	310

仓库 - 商品的销量

由于数据量太大，这里只展示一部分。

date	dc_id	sku_id	idx	sale
2018/5/1	DC001	SKU052	DC001_SKU052	0
2018/5/1	DC001	SKU060	DC001_SKU060	9
2018/5/1	DC001	SKU029	DC001_SKU029	589
2018/5/1	DC001	SKU036	DC001_SKU036	43
2018/5/1	DC001	SKU045	DC001_SKU045	2751
2018/5/1	DC001	SKU068	DC001_SKU068	96
2018/5/1	DC001	SKU069	DC001_SKU069	200
2018/5/1	DC001	SKU070	DC001_SKU070	42
2018/5/1	DC001	SKU071	DC001_SKU071	10
2018/5/1	DC001	SKU011	DC001_SKU011	12
2018/5/1	DC001	SKU013	DC001_SKU013	0
2018/5/1	DC001	SKU016	DC001_SKU016	19
2018/5/1	DC001	SKU046	DC001_SKU046	1417
2018/5/1	DC001	SKU047	DC001_SKU047	352
2018/5/1	DC001	SKU062	DC001_SKU062	678
2018/5/1	DC001	SKU063	DC001_SKU063	80
2018/5/1	DC001	SKU001	DC001_SKU001	0
2018/5/1	DC001	SKU002	DC001_SKU002	5
2018/5/1	DC001	SKU022	DC001_SKU022	1920

代码

导入包

Python

import pandas as pd
import networkx as nx
import matplotlib.pyplot as plt
from matplotlib_inline.backend_inline import set_matplotlib_formats

set_matplotlib_formats("svg")

工厂 - 仓库的运输成本

Python

def plot_transport_tarrif(self):
    """
    绘制 Factory-DC 运输成本网络图
    """
    # 创建一个空的无向图
    G = nx.Graph()
    # 向图中添加所有的工厂和配送中心
    G.add_nodes_from(self.factory_list + self.dc_list)
    # 向图中添加所有的运输费用记录，其中边的两端分别为配送中心
    G.add_edges_from(
        list(
            self.transport_tariff[["factory_id", "dc_id"]]
            .drop_duplicates()
            .itertuples(index=False, name=None)
        )
    )
    # 定义边的宽度，根据运输费用进行设置
    width = {}
    for idx, row in self.transport_tariff.iterrows():
        width[(row["factory_id"], row["dc_id"])] = row["unit_cost"] / 500
    # 获取一个二分图的布局
    pos = nx.bipartite_layout(G, self.factory_list)
    # 绘制图形，并设置边的宽度和节点的大小
    nx.draw(G, pos, width=[width[e] for e in G.edges], node_size=30)
    # 在左边添加“Factory”标注
    plt.text(-1.1, 0, "Factory", fontsize=12)
    # 在右边添加“DC”标注
    plt.text(0.15, 0, "DC", fontsize=12)
    # 添加标题
    plt.title("Transport Tariff")
    # 导出图像到本地
    plt.savefig("./image/transport_tariff.png", dpi=300, bbox_inches="tight")
    # 显示图像
    plt.show()

仓库 - 商品的销量

Python

def plot_dc_sku_network(self, date=None):
    """
    绘制 DC-SKU 销售量网络图

    Args:
        date: str “YYYY-MM-DD”, e.g. “2020-01-01”
    """
    if date:
        data = self.sales_data[self.sales_data.date == date]
        # 控制边的宽度
        reducer = 10000
    else:
        data = self.sales_data
        # 控制边的宽度
        reducer = 5000000
    # 创建一个空的无向图
    G = nx.Graph()
    # 向图中添加所有的配送中心和商品
    G.add_nodes_from(self.dc_list + self.sku_list)
    # 向图中添加所有的销售记录，其中边的两端分别为配送中心和商品
    G.add_edges_from(
        list(
            data[["dc_id", "sku_id"]]
            .drop_duplicates()
            .itertuples(index=False, name=None)
        )
    )
    # 定义边的宽度，根据销售量进行设置
    width = {}
    for idx, row in data.iterrows():
        width[(row["dc_id"], row["sku_id"])] = (
            width.get((row["dc_id"], row["sku_id"]), 0) + row["sale"] / reducer
        )
    # 获取一个二分图的布局
    pos = nx.bipartite_layout(G, self.dc_list)
    # 绘制图形，并设置边的宽度和节点的大小
    nx.draw(G, pos, width=[width[e] for e in G.edges], node_size=30)
    # 在左边添加“DC”标注
    plt.text(-1.1, 0, "DC", fontsize=12)
    # 在右边添加“SKU”标注
    plt.text(0.3, 0, "SKU", fontsize=12)
    # 添加标题，导出图像到本地
    if date:
        plt.title(f"DC-SKU Network {date}")
        plt.savefig(f"./image/dc_sku_network_{date}.png", dpi=300, bbox_inches="tight")
    else:
        plt.title("DC-SKU Network")
        plt.savefig("./image/dc_sku_network.png", dpi=300, bbox_inches="tight")
    # 显示图像
    plt.show()

评论