Uncovering Tianjin’s Bus Network: From Raw GPS Data to Complex Network Insights

1. Data Viewing and Preprocessing

Data obtained from the Gaode Map API includes bus line names, direction (0 for up, 1 for down), station order, station name, and longitude/latitude expressed in minutes. The original file contains 30,396 records; five station names, one latitude, and 38 longitude entries are missing and are removed for simplicity.

import pandas as pd

df = pd.read_excel('site_information.xlsx')
print(df.head())

Field description:

线路名称 – bus line name

上下行 – 0 for up, 1 for down

站序号 – station sequence number

站名称 – station name

经度（分） – longitude (minutes)

纬度（分） – latitude (minutes)

After cleaning, the dataset contains 6,618 bus lines and 4,851 stations.

df2 = df1.copy()
df2['经度（分）'] = df1['经度（分）'].apply(float) / 60
df2['纬度（分）'] = df1['纬度（分）'].apply(float) / 60
print(df2.head())

df2.to_excel('处理后数据.xlsx', index=False)

2. Data Analysis

Using matplotlib to plot a scatter diagram of longitude versus latitude reveals clear hotspots in Heping and Nankai districts.

# -*- coding: UTF-8 -*-
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib as mpl
import random

df = pd.read_excel('处理后数据.xlsx')
x_data = df['经度（分）']
y_data = df['纬度（分）']
colors = [random.choice(['#FF0000', '#0000CD', '#00BFFF', '#008000', '#FF1493', '#FFD700', '#FF4500', '#00FA9A', '#191970', '#9932CC']) for _ in range(len(x_data))]
mpl.rcParams['font.family'] = 'SimHei'
plt.style.use('ggplot')
plt.figure(figsize=(12, 6), dpi=200)
plt.scatter(x_data, y_data, marker='o', s=9., c=colors)
plt.xlabel('经度')
plt.ylabel('纬度')
plt.title('天津市公交站点分布情况')
plt.savefig('经纬度散点图.png')
plt.show()

Mapping the same data on an actual map with pyecharts BMap shows dense line networks in the same districts.

# -*- coding: UTF-8 -*-
import pandas as pd
from pyecharts.charts import BMap
from pyecharts import options as opts
from pyecharts.globals import CurrentConfig

CurrentConfig.ONLINE_HOST = 'D:/python/pyecharts-assets-master/assets/'

df = pd.read_excel('处理后数据.xlsx')
df.drop_duplicates(subset='站名称', inplace=True)
longitude = list(df['经度（分）'])
latitude = list(df['纬度（分）'])

datas = [list(zip(longitude, latitude))]
BAIDU_MAP_AK = '改成你的百度地图AK'

c = (BMap(init_opts=opts.InitOpts(width='1200px', height='800px'))
     .add_schema(baidu_ak=BAIDU_MAP_AK, center=[117.20, 39.13], zoom=10, is_roam=True)
     .add('', type_='lines', is_polyline=True, data_pair=datas,
          linestyle_opts=opts.LineStyleOpts(opacity=0.2, width=0.5, color='red'),
          progressive=200, progressive_threshold=500))

c.render('公交网络地图.html')

Network Degree Analysis

The degree of a line node equals the number of other lines reachable via a transfer. The network contains 618 lines; the maximum degree is 175, the minimum is 0, and the average degree is 55.41. Most degrees lie between 7 and 26.

# -*- coding: UTF-8 -*-
import xlrd
import pandas as pd
import collections
import matplotlib.pyplot as plt
import matplotlib as mpl

# Load raw data
df = pd.read_excel('site_information.xlsx')
line_names = df['线路名称'].unique()
line_list = list(line_names)

# Build station dictionary for each line (upward direction only)
data = xlrd.open_workbook('site_information.xlsx')
table = data.sheets()[0]
site_dic = {k: [] for k in line_list}
for i in range(1, table.nrows):
    row = table.row_values(i)
    if row[1] == '0':  # up direction
        site_dic[row[0]].append(row[3])

# Initialize degree list
node_count = [0] * len(line_list)
sites = list(site_dic.values())

# Compute degree by checking shared stations between lines
for j in range(len(sites)):
    for k in range(j + 1, len(sites)):
        if set(sites[j]) & set(sites[k]):
            node_count[j] += 1
            node_count[k] += 1

print(f"公交网络共有 {len(line_list)} 条线路")
print(f"线路网络的度的最大值为：{max(node_count)}")
print(f"线路网络的度的最小值为：{min(node_count)}")
print(f"线路网络的度的平均值为：{sum(node_count) / len(node_count)}")

# Plot degree distribution
node_number = list(range(len(node_count)))
mpl.rcParams['font.family'] = 'SimHei'
plt.figure(figsize=(10, 6), dpi=150)
plt.bar(node_number, node_count, color='purple')
plt.xlabel('节点编号n')
plt.ylabel('节点的度数K')
plt.title('线路网络中各节点的度的大小分布')
plt.savefig('线路网络中各节点的度的大小.png')
plt.show()

Clustering Coefficient

The clustering coefficient measures how tightly a node’s neighbors are connected. The average clustering coefficient of Tianjin’s bus network is 0.0906, lower than that of a comparable random network (≈0.00044), indicating a small‑world property.

# Compute clustering coefficient for each node
Ei = []  # actual number of edges among neighbors
for a in range(len(sites)):
    neighbor = []
    if node_count[a] <= 1:
        Ei.append(0)
        continue
    for b in range(len(sites)):
        if a == b:
            continue
        if set(sites[a]) & set(sites[b]):
            neighbor.append(b)
    # Count edges among neighbors
    count = 0
    for i in range(len(neighbor)):
        for j in range(i + 1, len(neighbor)):
            if set(sites[neighbor[i]]) & set(sites[neighbor[j]]):
                count += 1
    Ei.append(count)

# Clustering coefficient per node
Ci = []
for m in range(len(node_count)):
    if node_count[m] <= 1:
        Ci.append(0)
    else:
        Ci.append(2 * Ei[m] / (node_count[m] * (node_count[m] - 1)))

print("天津市公交线路网络平均聚类系数为：{:.4f}".format(sum(Ci) / len(Ci)))

# Plot clustering coefficient distribution
mpl.rcParams['font.family'] = 'SimHei'
plt.figure(figsize=(10, 6), dpi=150)
plt.bar(range(len(Ci)), Ci, color='blue')
plt.xlabel('节点编号n')
plt.ylabel('节点的聚类系数')
plt.title('线路网络中各节点的聚类系数分布')
plt.savefig('聚类系数分布.png')
plt.show()

Reference: "Complex Network Analysis of Tianjin Public Transport" and related studies on urban bus network topology.