R语言Keras训练神经网络报错：输入数据不匹配求助

使用R语言结合Keras和TensorFlow训练神经网络时，调用fit方法触发以下报错：

Error in py_call_impl(callable, call_args$unnamed, call_args$named) :
ValueError: in user code:

<...truncated...>/site-packages/keras/src/engine/training.py", line 1080, in train_step
y_pred = self(x, training=True)
File "/Users/ronypabraham/.virtualenvs/r-tensorflow/lib/python3.9/site-packages/keras/src/utils/traceback_utils.py", line 70, in error_handler
raise e.with_traceback(filtered_tb) from None
File "/Users/ronypabraham/.virtualenvs/r-tensorflow/lib/python3.9/site-packages/keras/src/engine/input_spec.py", line 197, in assert_input_compatibility
raise ValueError(

    ValueError: Missing data for input "dense_39_input". You passed a data dictionary with keys ['VehPower', 'VehAge', 'DrivAge', 'BonusMalus', 'VehBrand_embedded', 'VehGas_encoded', 'Region_embedded', 'Area_encoded', 'Density']. Expected the following keys: ['dense_39_input']

触发报错的fit代码：

history <- model %>%
fit(
x = X_train,
y = y_train,
batch_size = 8,
epochs = 50,
verbose = 2,
validation_split = 0.2
)

1. 模型构建错误：添加层时使用管道操作但未将结果赋值回model对象，导致模型结构未正确初始化。
2. 输入数据格式不匹配：X_train为data.frame类型，Keras会将其解析为以列名为键的输入字典，但Sequential模型期望单一张量输入，输入键为默认的层输入名（如dense_39_input），因此出现键不匹配。

1. 正确构建模型 #

将层添加操作通过链式管道赋值给model，确保层被正确添加到模型中：

# Create a simple feedforward neural network model
model <- keras_model_sequential() %>%
  # Add input layer
  layer_dense(units = 20, activation = "relu", input_shape = 9) %>% # 9 features
  # Add output layer for count data
  layer_dense(units = 1, activation = "linear") # 适用于计数数据

2. 转换输入数据格式 #

将X_train和X_test从data.frame转换为矩阵，避免Keras将其解析为字典输入：

# Predictor variables
X_train <- as.matrix(train_data[, c("VehPower", "VehAge", "DrivAge", "BonusMalus", 
                                    "VehBrand_embedded", "VehGas_encoded", "Region_embedded",  "Area_encoded", "Density")])

X_test <- as.matrix(test_data[, c("VehPower", "VehAge", "DrivAge", "BonusMalus", 
                                  "VehBrand_embedded", "VehGas_encoded", "Region_embedded", "Area_encoded", "Density")])

3. 优化模型编译（可选） #

计数数据回归任务中，accuracy指标不适用，建议替换为回归类指标：

# Compile the model
model$compile(
  loss = "poisson",
  optimizer = optimizer,
  metrics = c('mean_absolute_error')
)

运行summary(model)确认模型结构正确，输入形状应为(None, 9)，输出层形状为(None, 1)，之后重新执行fit方法即可正常训练。

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