Keras模型输入维度不匹配问题解析与解决方案

# 在模型定义前验证X_train的维度
print("X_train shape:", X_train.shape)
print("X_train columns:", X_train.columns)

# ... 模型定义 ...

# 在模型拟合前再次验证X_train的维度
print("X_train shape before fit:", X_train.shape)

import pandas as pd
from sklearn.model_selection import train_test_split
from keras.models import Sequential
from keras.layers import Dense, Dropout
from keras.optimizers import Adam
from keras.regularizers import l2
import numpy as np

# 假设 dataset.csv 包含 'Superficie', 'Numero di stanze da letto', 'Numero di bagni', 'Anno di costruzione', 'Località', 'Prezzo'
def carica_dataset():
    dataset = pd.read_csv("dataset.csv")
    return dataset

def crea_并训练_模型():
    dataset = carica_dataset()

    # 存储训练数据集的原始列名，以便后续对齐
    original_columns = dataset.drop(columns=['Prezzo']).columns

    # 对训练数据进行独热编码
    # 记录独热编码后的所有列名，这是模型期望的输入特征
    dataset_encoded = pd.get_dummies(dataset, columns=['Località'], drop_first=False) 

    X = dataset_encoded.drop(columns=['Prezzo'])
    y = dataset_encoded['Prezzo']

    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42) # 增加test_size和random_state

    # 打印X_train的维度，确保 input_dim 设置正确
    print(f"训练集特征维度 X_train.shape[1]: {X_train.shape[1]}")

    model = Sequential()
    model.add(Dense(64, activation='relu', input_dim=X_train.shape[1],  kernel_regularizer=l2(0.1)))
    model.add(Dropout(0.5))
    model.add(Dense(32, activation='relu',  kernel_regularizer=l2(0.1)))
    model.add(Dropout(0.5))
    model.add(Dense(16, activation='relu', kernel_regularizer=l2(0.1)))
    model.add(Dropout(0.5))
    model.add(Dense(8, activation='relu', kernel_regularizer=l2(0.1)))
    model.add(Dropout(0.5))
    model.add(Dense(1, activation='linear', kernel_regularizer=l2(0.1)))

    adam = Adam(learning_rate=0.001) # 明确指定学习率
    model.compile(loss='mean_squared_error', optimizer=adam, metrics=['mse']) # metrics改为mse更合理

    model.fit(X_train, y_train, epochs=100, batch_size=64, verbose=0) # verbose=0 减少输出

    return model, X_train.columns # 返回训练时使用的列名

# 主程序逻辑
model, train_columns = crea_并训练_模型()

fields = {
    'Superficie': float,
    'Numero di stanze da letto': int,
    'Numero di bagni': int,
    'Anno di costruzione': int,
    'Località': str
}
user_data = {}

print("\n请输入预测房屋信息：")
for key, value in fields.items():
    while True:
        try:
            user_input = input(f"请输入 {key} 的值: ")
            user_data[key] = value(user_input)
            break
        except ValueError:
            print(f"输入无效，请为 {key} 输入一个有效值。")

# 将用户输入转换为DataFrame
dataframe_user = pd.DataFrame([user_data])

# 对用户输入数据进行独热编码
dataframe_user_encoded = pd.get_dummies(dataframe_user, columns=['Località'], drop_first=False)

# 关键步骤：使用训练时的列名对用户输入数据进行reindex
# 这将确保预测数据的列数和顺序与训练数据完全一致
# 如果有训练时存在的列在当前用户数据中不存在，它将被填充为0
# 如果用户数据中存在训练时不存在的Località，则该列会被忽略（因为不在train_columns中）
prediction_input = dataframe_user_encoded.reindex(columns=train_columns, fill_value=0)

# 打印预测输入数据的形状和列名，进行验证
print(f"\n预测输入数据形状: {prediction_input.shape}")
print(f"预测输入数据列名: {prediction_input.columns.tolist()}")

# 确保所有列都是数值类型
valori = prediction_input.values

# 进行预测
prediction = model.predict(valori)[0][0]
print(f'\n预测的房屋价格是: {prediction:.2f} €')