金融工程--基础概念

背景

1、自学金融可能遇到的问题。
2、搭建量化所需的框架探索。
3、策略探索。

框架对比

Aspect	First Script	Second Script	Third Script
Data Source	AkShare API	AkShare API	AkShare API
Models Used	SVC	Logistic Regression, GaussianNB, SVC, Random Forest, MLP	Logistic Regression, GaussianNB, SVC, Random Forest, MLP
Feature Engineering	Lagged returns, direction as features	Lagged returns, direction as features	Lagged returns, direction as features
Evaluation Metrics	Accuracy Score, Classification Report, Confusion Matrix	Total Returns, Annual Volatility, Accuracy Score	Total Returns, Annual Volatility, Accuracy Score
Backtesting Strategy	Uses predictions to decide full buy/sell strategy, no shorting	Uses multiple models for predictions, reports on highest accuracy	Uses best-performing model (SVM) for backtest strategy
Performance Visualization	Performance comparison with benchmark using Pyfolio	Uses Pyfolio for detailed strategy evaluation and plotting	Detailed plotting using Pyfolio for various metrics
Code Organization	Functions for data download, feature creation, train/test split	More modular with functions for models and evaluation	Highly modular, structured with dedicated functions
Key Libraries	AkShare, sklearn, backtrader, pyfolio	AkShare, sklearn, backtrader, pyfolio	AkShare, sklearn, backtrader, pyfolio
Handling of NaN Values	Dropped NaNs before feature creation	Filled NaNs with 0 before conversion to int	Filled NaNs with 0 before conversion to int

基础概念

股票价格进行前复权和后复权的计算。在量化分析中，后复权数据由于其连续性，更适合用来计算股票的真实收益率和进行长期投资分析。前复权数据虽然方便观察当前价格与历史价格的关系，但可能因分红送股等事件导致历史价格出现负值，因此在量化策略中使用较少。

import pandas as pd

# 假设我们有以下DataFrame，包含日期、原始收盘价、每股分红和每股转送股数
data = {
    'date': ['2024-08-01', '2024-08-02', '2024-08-03'],
    'close_price': [100, 90, 95],  # 原始收盘价
    'dividend_per_share': [0, 2, 0],  # 每股分红
    'share_transfer_per_share': [0, 0, 0.5]  # 每股转送股数
}
df = pd.DataFrame(data)
df['date'] = pd.to_datetime(df['date'])

# 计算前复权因子
df['adjust_factor'] = (1 + df['share_transfer_per_share']) * (1 - df['dividend_per_share'] / df['close_price'])

# 计算前复权价格
df['forward_adjusted_price'] = df['close_price'] / df['adjust_factor'].shift(1).fillna(1)

# 计算后复权价格
df['backward_adjusted_price'] = df['close_price'] * df['adjust_factor'].cumprod()

# 显示结果
print(df)

import akshare as ak
import pandas as pd

# 获取上证指数的历史数据，指定后复权
stock_zh_a_hist_df = ak.stock_zh_a_hist(symbol='000001.SH', 
                                       start_date='20200101', 
                                       end_date='20240829', 
                                       adjust='hfq')

# 将数据转换为DataFrame，并设置日期列为索引
df = pd.DataFrame(stock_zh_a_hist_df['日K线'])
df.index = pd.to_datetime(df['日期'])
df.drop('日期', axis=1, inplace=True)

# 显示数据
print(df.head())

# 如果需要前复权数据，只需将adjust参数改为'qfq'
# stock_zh_a_hist_df = ak.stock_zh_a_hist(symbol='000001.SH', adjust='qfq')

参考资料

https://akshare.akfamily.xyz/introduction.html
https://github.com/akfamily/akshare