以下是另一种利用深度学习实现优质股票推送通知的示例代码，这里使用 pandas-datareader 库获取股票数据，以简化数据获取过程，并在模型构建上稍作调整：
import pandas as pd
import numpy as np
from pandas_datareader import data as pdr
import yfinance as yf
yf.pdr_override() # 覆盖 pandas_datareader 的默认数据源为 yfinance
from sklearn.preprocessing import MinMaxScaler
from keras.models import Sequential
from keras.layers import GRU, Dense, Dropout
from keras.callbacks import EarlyStopping
import matplotlib.pyplot as plt
import telegram # 推送通知所需库，提前安装# 示例股票代码列表，可按需修改或扩展
stock_symbols = ['TSLA', 'NVDA', 'JPM', 'BAC']# 1. 数据获取与处理
def get_stock_info(symbol, start_date, end_date):"""获取指定股票在给定时间段内的数据"""try:stock_data = pdr.get_data_yahoo(symbol, start=start_date, end=end_date)return stock_dataexcept Exception as e:print(f"Error fetching data for {symbol}: {e}")return None# 存储各股票数据的字典
stock_data_dict = {}
for symbol in stock_symbols:data = get_stock_info(symbol, '2015-01-01', '2022-12-31')if data is not None:stock_data_dict[symbol] = data# 合并数据，提取特征与目标变量，并归一化特征
combined_df = pd.DataFrame()
for symbol, data in stock_data_dict.items():data['Stock'] = symbolcombined_df = combined_df.append(data)features = combined_df[['Open', 'High', 'Low', 'Volume']]
target = combined_df['Close']scaler = MinMaxScaler()
scaled_features = scaler.fit_transform(features)# 划分训练集和测试集
train_size = int(len(scaled_features) * 0.8)
train_features = scaled_features[:train_size]
test_features = scaled_features[train_size:]
train_target = target[:train_size]
test_target = target[train_size:]# 重塑数据以适配 GRU 模型输入
def reshape_for_gru(data, timesteps):"""将数据重塑为适合 GRU 模型的格式"""X, y = [], []for i in range(len(data) - timesteps):X.append(data[i:i + timesteps])y.append(data[i + timesteps])return np.array(X), np.array(y)timesteps = 20 # 时间步长，可按需调整
train_X, train_y = reshape_for_gru(train_features, timesteps)
test_X, test_y = reshape_for_gru(test_features, timesteps)# 2. 模型构建
model = Sequential()
model.add(GRU(units=60, return_sequences=True, input_shape=(timesteps, train_X.shape[2])))
model.add(Dropout(0.2))
model.add(GRU(units=60))
model.add(Dropout(0.2))
model.add(Dense(units=1))# 3. 模型训练
model.compile(optimizer='adam', loss='mean_squared_error')
early_stop = EarlyStopping(monitor='val_loss', patience=8, verbose=1)
model.fit(train_X, train_y, epochs=80, batch_size=32, validation_split=0.2, callbacks=[early_stop])# 4. 模型评估
train_pred = model.predict(train_X)
test_pred = model.predict(test_X)# 反归一化预测结果
train_pred = scaler.inverse_transform(train_pred)
test_pred = scaler.inverse_transform(test_pred)# 计算评估指标，此处以均方根误差为例
from sklearn.metrics import mean_squared_error
train_rmse = np.sqrt(mean_squared_error(train_y, train_pred))
test_rmse = np.sqrt(mean_squared_error(test_y, test_pred))
print(f"Train RMSE: {train_rmse}")
print(f"Test RMSE: {test_rmse}")# 5. 预测未来走势并推送通知
def predict_future_stock(symbol, model, scaler, timesteps, days):"""预测指定股票未来几天的走势"""last_data = stock_data_dict[symbol].tail(timesteps)last_features = last_data[['Open', 'High', 'Low', 'Volume']]scaled_last_features = scaler.transform(last_features)future_preds = []for _ in range(days):pred = model.predict(scaled_last_features[-timesteps:].reshape(1, timesteps, 4))future_preds.append(pred)scaled_last_features = np.vstack([scaled_last_features[1:], pred])return scaler.inverse_transform(np.array(future_preds).squeeze())# 预测未来 3 天走势，天数可按需调整
future_days = 3
for symbol in stock_symbols:future_prediction = predict_future_stock(symbol, model, scaler, timesteps, future_days)print(f"Future {future_days} days prediction for {symbol}:")print(future_prediction)# 简单判断涨幅，设涨幅超 3%为优质股票（可调整）current_price = stock_data_dict[symbol]['Close'].iloc[-1]for i in range(future_days):if (future_prediction[i] / current_price - 1) > 0.03:bot_token = 'YOUR_BOT_TOKEN'chat_id = 'YOUR_CHAT_ID'bot = telegram.Bot(token=bot_token)message = f"{symbol} may have a good increase in the next {i + 1} days. Current price: {current_price}, Predicted price: {future_prediction[i]}"bot.sendMessage(chat_id=chat_id, text=message)请注意：
- 代码中的
YOUR_BOT_TOKEN和YOUR_CHAT_ID要替换为你通过 Telegram Bot API 申请的真实信息，才能实现推送通知。 - 股票市场受诸多复杂因素影响，深度学习模型仅基于历史数据预测，仅供参考，不能完全等同于未来真实走势。
- 模型架构、参数（如
GRU单元数、时间步长等）以及数据处理步骤都可依据实际情况进一步优化，以适应不同股票数据特性与预测要求。
实际运用时，建议持续优化数据处理、模型搭建及验证流程，融入更多金融专业知识与实时市场信息，提升预测的精准度与可靠性。例如，可考虑加入宏观经济指标作为特征，或采用更复杂的模型融合策略来增强预测能力。