2024HuoHua C题第三问

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from statsmodels.tsa.statespace.sarimax import SARIMAX


plt.rcParams['font.sans-serif'] = ['SimHei']    # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False    # 用来显示负号

pd.set_option('display.max_columns', None)    # 显示所有列
pd.set_option('display.max_rows', None)      # 显示所有行
pd.set_option('max_colwidth', 400)


# 定义一个使用SARIMAX模型预测的函数
def predict_sarimax(series, exog, steps=7):
    model = SARIMAX(series, order=(5, 1, 0), exog=exog)
    model_fit = model.fit(disp=False)
    forecast = model_fit.forecast(steps=steps, exog=exog[-steps:])
    return forecast


# 重新加载 CSV 文件
df_past_order = pd.read_csv("C:\\Users\\ChenCong\\Desktop\\2024年度“火花杯”数学建模精英联赛-C题-附件\\2024年度“火花杯”数学建模精英联赛-C题-附件\\input_data\\df_past_order.csv")
df_loc = pd.read_csv("C:\\Users\\ChenCong\\Desktop\\2024年度“火花杯”数学建模精英联赛-C题-附件\\2024年度“火花杯”数学建模精英联赛-C题-附件\\input_data\\df_loc.csv")

# # 去除 'Name' 列中末尾的 'shi' 和 '-'
# df_past_order['Name'] = df_past_order['Name'].str.replace('shi$', '', regex=True).str.rstrip('-')
df_loc['name'] = df_loc['name'] + '-shi'

# 重新合并数据集
df_merged_corrected = pd.merge(df_past_order, df_loc, left_on='Name', right_on='name', how='left')

# print(df_merged_corrected)
# print('-------------------------------------------------------')

# 删除冗余的 'name' 列
df_merged_corrected.drop(columns=['name'], inplace=True)

date_columns = df_past_order.columns[2:]
print(date_columns)

df_merged_corrected[['Longitude', 'Latitude', 'resident_pop', 'gdp']] = df_merged_corrected[['Longitude', 'Latitude', 'resident_pop', 'gdp']].apply(pd.to_numeric, errors='coerce')
df_merged_corrected.fillna(0, inplace=True)  # 可以使用0填充NaN，或者根据需求使用其他方法

# # 查看数据框中每列的数据类型
# print(df_merged_corrected.dtypes)
# print('---------------------------------------------------')
# # 检查是否有缺失值（NaN）
# print(df_merged_corrected[['Longitude', 'Latitude', 'resident_pop', 'gdp']].isnull())
# print(df_merged_corrected.isnull().sum())
# print('---------------------------------------------------')

# df_merged_corrected.to_excel('df_merged_corrected.xlsx')

# 遍历每个城市和SKU进行预测
predictions = []

for _, row in df_merged_corrected.iterrows():
    city = row['Name']
    sku = row['SKU']
    series = row[date_columns].values.astype(float)

    # 打印要传递给模型的数据
    print("Exogenous variables (exog):", row[['Longitude', 'Latitude', 'resident_pop', 'gdp']].values)
    print("Time series data (series):", row[date_columns].values)

    # 选择外生变量（例如，地理信息、人口和GDP）
    exog = row[['Longitude', 'Latitude', 'resident_pop', 'gdp']].values.reshape(1, -1)

    print(exog)

    # 预测未来7天的订单量
    forecast = predict_sarimax(series, exog, steps=7)

    # 将结果添加到预测列表中
    prediction = {'Name': city, 'SKU': sku}
    for i, val in enumerate(forecast):
        prediction[f'forecast_day_{i + 1}'] = val
    predictions.append(prediction)

# 将预测结果转换为DataFrame格式
df_predictions = pd.DataFrame(predictions)

# 绘制每个城市的未来7天预测数据

# 设置图形大小
plt.figure(figsize=(12, 8))

# 遍历每个城市的预测数据
for i, row in df_predictions.iterrows():
    days = [f'forecast_day_{j + 1}' for j in range(7)]  # 创建未来7天的标签
    values = row[days].values  # 获取对应的预测值

    plt.plot(days, values, marker='o', label=f'{row["Name"]} ({row["SKU"]})')

    if i == 5:
        break

# 添加标题和标签
plt.title('未来7天的订单预测')
plt.xlabel('预测日期')
plt.ylabel('预测订单量')
plt.legend(loc='upper right')
plt.grid(True)

# 显示图形
plt.show()

错误提示：

Traceback (most recent call last):
  File "E:\Catch_pigs\23C\test3.py", line 74, in <module>
    forecast = predict_sarimax(series, exog, steps=7)
  File "E:\Catch_pigs\23C\test3.py", line 17, in predict_sarimax
    model = SARIMAX(series, order=(5, 1, 0), exog=exog)
  File "E:\python test\venv\lib\site-packages\statsmodels\tsa\statespace\sarimax.py", line 328, in __init__
    self._spec = SARIMAXSpecification(
  File "E:\python test\venv\lib\site-packages\statsmodels\tsa\arima\specification.py", line 446, in __init__
    self._model = TimeSeriesModel(endog, exog=exog, dates=dates, freq=freq,
  File "E:\python test\venv\lib\site-packages\statsmodels\tsa\base\tsa_model.py", line 470, in __init__
    super().__init__(endog, exog, missing=missing, **kwargs)
  File "E:\python test\venv\lib\site-packages\statsmodels\base\model.py", line 270, in __init__
    super().__init__(endog, exog, **kwargs)
  File "E:\python test\venv\lib\site-packages\statsmodels\base\model.py", line 95, in __init__
    self.data = self._handle_data(endog, exog, missing, hasconst,
  File "E:\python test\venv\lib\site-packages\statsmodels\base\model.py", line 135, in _handle_data
    data = handle_data(endog, exog, missing, hasconst, **kwargs)
  File "E:\python test\venv\lib\site-packages\statsmodels\base\data.py", line 675, in handle_data
    return klass(endog, exog=exog, missing=missing, hasconst=hasconst,
  File "E:\python test\venv\lib\site-packages\statsmodels\base\data.py", line 88, in __init__
    self._handle_constant(hasconst)
  File "E:\python test\venv\lib\site-packages\statsmodels\base\data.py", line 133, in _handle_constant
    if not np.isfinite(exog_max).all():
TypeError: ufunc 'isfinite' not supported for the input types, and the inputs could not be safely coerced to any supported types according to the casting rule ''safe''

将错误提示传给ChatGPT：

经过测试发现，输入只包含float64并且没有NaN或者inf，因此问题并没有找到。

可能原因是模型自身问题。