Keras模型输入维度不匹配：数据预处理一致性指南

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

def carica_dataset():
    dataset = pd.read_csv("dataset.csv")
    return dataset

# 全局变量或通过函数返回，用于存储训练时的特征列
training_feature_columns = None 

def carica_modello():
    global training_feature_columns # 声明使用全局变量
    dataset = carica_dataset()

    # 对整个数据集进行独热编码，确保所有可能的类别都被考虑
    # 或者只对训练集进行独热编码，然后将编码器保存下来
    # 这里我们选择在训练前对整个数据集进行独热编码，以获取所有可能的列
    # 但更严谨的做法是先分训练集测试集，然后对训练集进行get_dummies，并保存列名

    # 假设我们只对训练集进行get_dummies，并保存列名
    X_raw = dataset.drop(columns=['Prezzo'])
    y = dataset['Prezzo']

    X_train_raw, X_test_raw, y_train, y_test = train_test_split(X_raw, y, test_size=0.2, random_state=42)

    # 对训练集进行独热编码
    X_train = pd.get_dummies(X_train_raw, columns=['Località'])
    # 记录训练集独热编码后的列名
    training_feature_columns = X_train.columns.tolist() 

    # 对测试集进行独热编码，并确保列与训练集一致
    X_test = pd.get_dummies(X_test_raw, columns=['Località'])
    X_test = X_test.reindex(columns=training_feature_columns, fill_value=0) # 填充缺失列为0

    print(f"X_train shape after one-hot encoding: {X_train.shape}")
    print(f"X_train columns: {X_train.columns.tolist()}")

    model = Sequential()
    # input_dim 必须与 X_train 的特征数量一致
    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()

    model.compile(loss='mean_squared_error', optimizer=adam, metrics=['accuracy'])

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

    return model

# 主程序部分
dataset = carica_dataset()
model = carica_modello() # 此时 training_feature_columns 已经被填充

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

for key,value in fields.items():
    while True:
        try:
            user_input = input(f"inserisci il valore di: {key}: ")
            user_data[key] = value(user_input)
            break
        except ValueError:
            print(f"inserisci un valore valido per {key}")

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

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

# 关键步骤：使用训练时的列名对用户数据DataFrame进行reindex
# 这将确保用户数据DataFrame拥有与X_train完全相同的列，缺失的列填充为0
if training_feature_columns is not None:
    dataframe = dataframe.reindex(columns=training_feature_columns, fill_value=0)
else:
    raise ValueError("Training feature columns not set. Ensure carica_modello was called.")

print(f"User input dataframe shape after reindexing: {dataframe.shape}")
print(f"User input dataframe columns: {dataframe.columns.tolist()}")

valori = dataframe.values

prediction = model.predict(valori)[0][0]
print(f'La predizione del prezzo è: {prediction} €')