使用回归模型预测时遇ValueError：无法将'veh2312'转为整数

问题描述 #

使用随机森林回归模型做预测时触发错误：ValueError: invalid literal for int() with base 10: 'veh2312'，多次尝试修复无效，寻求解决方法。

错误栈 #

---> 47 sample_input = df[(df['vehicle_id'] == label_encoder.transform([vehicle_id])[0]) & (df['time'] == time_step)]
/usr/local/lib/python3.10/dist-packages/sklearn/utils/_array_api.py in astype(self, x, dtype, copy, casting)
     66     def astype(self, x, dtype, *, copy=True, casting="unsafe"):
     67         # astype is not defined in the top level NumPy namespace
---> 68         return x.astype(dtype, copy=copy, casting=casting
ValueError: invalid literal for int() with base 10: 'veh2312'

完整代码 #

#'time' column to Unix timestamp
df['time'] = pd.to_datetime(df['time']).astype(int) // 10**9
label_encoder = LabelEncoder()
df['vehicle_id'] = label_encoder.fit_transform(df['vehicle_id'])  
X = df[['time', 'vehicle_id', 'vehicle_lat', 'vehicle_lon', 'cell_lat', 'cell_lon']]  # Features
y = df.groupby(['time', 'vehicle_id'])['cell_id'].nunique().reset_index(name='cell_id_count')
df_merged = pd.merge(df, y, on=['time', 'vehicle_id'])

y = df_merged['cell_id_count']
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
rf_regressor = RandomForestRegressor(n_estimators=10, random_state=42)
rf_regressor.fit(X_train, y_train)
y_pred = rf_regressor.predict(X_test)


#input
vehicle_id = 'veh2312'
time_step = 2194

#filter DataFrame based on 'vehicle_id' and 'time'
sample_input = df[(df['vehicle_id'] == label_encoder.transform([vehicle_id])[0]) & (df['time'] == time_step)]

if len(sample_input) > 0:
    sample_features = sample_input[['time', 'vehicle_id', 'vehicle_lat', 'vehicle_lon', 'cell_lat', 'cell_lon']]
    
    sample_predictions = rf_regressor.predict(sample_features)
    
    print(f"Predicted number of cell_ids for vehicle '{vehicle_id}' at time step {time_step}: {sample_predictions[0]:.0f}")
else:
    print(f"No data found for vehicle '{vehicle_id}' at time step {time_step}.")

样本数据集 #

cell_id cell_lat    cell_lon    time    vehicle_lat vehicle_lon vehicle_id  distance
2665    43.720118   10.422051   2194    43.70994621 10.42621903 veh2312 1205
2682    43.695238   10.433212   2194    43.70994621 10.42621903 veh2312 1787
20020711    43.716044   10.430057   2194    43.70994621 10.42621903 veh2312 791
27025350    43.710199   10.4247325  2194    43.70994621 10.42621903 veh2312 167
118817885   43.712409   10.424214   2194    43.70994621 10.42621903 veh2312 349
118825175   43.706199   10.431691   2194    43.70994621 10.42621903 veh2312 731
118840263   43.706233   10.432086   2194    43.70994621 10.42621903 veh2312 766

解决方案 #

错误原因 #

1. 时间列处理逻辑错误：代码错误地将代表时间步长的time列转换为Unix时间戳，导致df['time']的值与输入的time_step=2194完全不匹配；若数据读取时存在列错位，会导致time列混入vehicle_id的字符串值。
2. 数据类型不匹配：当df['time']包含字符串（如'veh2312'）时，执行df['time'] == time_step（整数）时，Pandas会尝试将字符串转换为整数，触发ValueError。

修复步骤 #

1. 修正时间列处理：如果time列是时间步长而非日期时间，删除错误的Unix时间戳转换代码，保留原始数值：

   # 移除该行错误代码
   # df['time'] = pd.to_datetime(df['time']).astype(int) // 10**9
   

2. 确保数据读取正确：读取数据时指定正确的分隔符和表头，保证time列为整数类型，vehicle_id列为字符串类型：

   df = pd.read_csv('your_data.csv', sep='\s+', header=0)
   # 强制转换time列为整数
   df['time'] = df['time'].astype(int)
   

3. 验证筛选逻辑：筛选样本前确认数据类型一致，避免隐式转换错误：

   encoded_vehicle_id = label_encoder.transform([vehicle_id])[0]
   # 确保time_step与df['time']类型匹配
   sample_input = df[(df['vehicle_id'] == encoded_vehicle_id) & (df['time'] == int(time_step))]
   

内容的提问来源于stack exchange，提问作者user1566490