Python地铁站点客流量预测：随机森林极限梯度提升回归器XGBoost|数据分享

import pandas as pd

data=pd.read_excel(‘总数据预测.xlsx’)

x=data.iloc[:,:5] #提取前四列数据

y=data.iloc[:,5] # 人流量数据

import numpy as py

x11=np.array([121,14967,12260,20151201,0])

…

x207=np.array([159,14132,14167,20151207,0])

x11=x11.reshape(1,5)

…

X207=x207.reshape(1,5)

#导入神经网络回归模块MLPRegressor。

from sklearn.neural_network import MLPRegressor

#利用MLPRegressor创建神经网络回归对象clf

Clf=MLPRegressor(solver=’lbfgs’,alpha=1e-5,hidden_layer_sizes=8,random_state=1)

#参数说明：

#solver:神经网络优化求解算法

#alpha:模型训练误差，默认为0.00001

#hidden_layer_sizes：隐含层神经元个数

#random_state：默认设置为1 #用clf对象中的fit()方法进行网络训练

clf.fit(x,y) #调用clf对象中的score()方法，获得神经网络回归的拟合优度（判决系数）

rv=clf.score(x,y) #调用clf对象中的predict()可以对测试样本进行预测，获得其测试结果

R11=clf.predict(x11)

R207=clf. predict (x207)

python复制import pandas as pd

# 加载CSV文件
df = pd.read_csv(file_path)

# 显示前几行数据以了解其结构
df.head()

python复制import matplotlib.pyplot as plt

# 将date_time列转换为datetime类型
df['date_time'] = pd.to_datetime(df['date_time'])

# 对数据进行排序以确保时间顺序
df = df.sort_values('date_time')

# 绘制折线图
plt.figure(figsize=(15, 6))
plt.plot(df['date_time'], df['traffic_volume'], marker='o', linestyle='-')
plt.title('24小时内交通流量变化趋势')
plt.xlabel('时间')
plt.ylabel('交通流量')
plt.grid(True)
plt.xticks(rotation=45)
plt.tight_layout()
plt.show()

python复制# 筛选出特定日期的数据（例如最早的日期）
specific_date = df['date_time'].dt.date.iloc[0]
df_specific_date = df[df['date_time'].dt.date == specific_date]

# 对数据进行排序以确保时间顺序
df_specific_date = df_specific_date.sort_values('date_time')

# 绘制折线图
plt.figure(figsize=(15, 6))
plt.plot(df_specific_date['date_time'], df_specific_date['traffic_volume'], marker='o', linestyle='-')
plt.title(f'{specific_date} 交通流量变化趋势')
plt.xlabel('时间')
plt.ylabel('交通流量')
plt.grid(True)
plt.xticks(rotation=45)
plt.tight_layout()
plt.show()

python复制import seaborn as sns

# 选择相关的列
columns_to_explore = ['traffic_volume', 'weather_main', 'temp', 'rain_1h', 'snow_1h', 'clouds_all']

# 由于weather_main是分类变量，我们需要将其转换为数值编码
df['weather_main_code'] = df['weather_main'].astype('category').cat.codes

# 创建散点图矩阵
sns.pairplot(df[columns_to_explore], diag_kind='kde', markers=["o", "s", "D", "v", "<", ">"])
plt.suptitle('交通流量与天气条件的关系', y=1.02)
plt.show()

python复制# 筛选出特定的假日
specific_holidays = ['Columbus Day', 'Veterans Day', 'Thanksgiving Day']
df_holidays = df[df['holiday'].isin(specific_holidays) | (df['holiday'] == 'None')]

# 创建箱线图
plt.figure(figsize=(12, 6))
sns.boxplot(x='holiday', y='traffic_volume', data=df_holidays, palette=["rgb(202,206,210, max =255)", "rgb(230,232,234, max =255)"])
plt.title('假日与非假日交通流量对比')
plt.xlabel('假日类型')
plt.ylabel('交通流量')
plt.show()

python复制# 修正颜色格式并重新创建箱线图
plt.figure(figsize=(12, 6))
sns.boxplot(x='holiday', y='traffic_volume', data=df_holidays, palette=["#C0C4C8", "#E6E8EA"])
plt.title('假日与非假日交通流量对比')
plt.xlabel('假日类型')
plt.ylabel('交通流量')
plt.show()

首先，我们对特征集`X`和目标变量`y`进行了训练集和测试集的划分，其中测试集占比为25%，并确保了数据的随机性，以避免任何潜在的偏差。具体实现如下：

                                                 y,
                                                 test_size=0.25,
                                                 shuffle=True,
                                                 random_state=42)

                 verbosity=2,
                 n_jobs=-1)

y_predictions = tpot.predict(X_test)
r2 = sklearn.metrics.r2_score(y_test, y_predictions)
mae = sklearn.metrics.mean_absolute_error(y_test, y_predictions)
mse = sklearn.metrics.mean_squared_error(y_test, y_predictions)
rmse = np.sqrt(mse)

training_features, testing_features, training_target, testing_target = \
            train_test_split(features, tpot_data['traffic_volume'].values, random_state=None)

exported_pipeline = make_pipeline(
    StandardScaler(),
    RandomForestRegressor(bootstrap=False, max_features=0.45, min_samples_leaf=2, min_samples_split=2, n_estimators=100)
)

exported_pipeline = RandomForestRegressor(bootstrap=True, max_features=0.9000000000000001, min_samples_leaf=9, min_samples_split=19, n_estimators=100)

exported_pipeline.fit(training_features, training_target)
results = exported_pipeline.predict(testing_features)

features = tpot_data.drop('traffic_volume', axis=1).values
training_features, testing_features, training_target, testing_target = \
            train_test_split(features, tpot_data['traffic_volume'].values, random_state=None)

exported_pipeline = ExtraTreesRegressor(bootstrap=False, max_features=0.8, min_samples_leaf=8, min_samples_split=18, n_estimators=100)

exported_pipeline.fit(training_features, training_target)
results = exported_pipeline.predict(testing_features)

a=tpot_data['traffic_volume'].values

features = tpot_data.drop('traffic_volume', axis=1).values
training_features, testing_features, training_target, testing_target = \
            train_test_split(features, tpot_data['traffic_volume'].values, random_state=None)

exported_pipeline = XGBRegressor(learning_rate=0.1, max_depth=3, min_child_weight=15, n_estimators=100, nthread=1, subsample=0.6000000000000001,objective = 'reg:squarederror')

exported_pipeline.fit(training_features, training_target)
results = exported_pipeline.predict(testing_features)