基于SVM与现代投资组合理论的量化框架

# QT_Test 本测试旨在重现一套比较简单且完备的量化框架，该框架基于现代投资组合理论，并应用主流的机器学习算法（SVM）进行分析。 旨在初步形成一个量化投资的思路，辅助构建科学合理的投资策略。 ## Anticipate Process ### Preparation - [SQL Queries](https://github.com/FDUJiaG/QT_Test/tree/master/sqlQueries) - [Initial Capital](https://github.com/FDUJiaG/QT_Test/blob/master/sqlQueries/my_capital.sql) of Loopback Test (optional, default = 1 M) ### Input - Stock Pool - Base Stock Index - Interval of Loopback Test - Windows of Preproession (optional, default = 365) - Windows of Loopback Trainning Test (optional, default = 90) - Windows of Loopack Portfolio (optional, default = year) - Change Frequency of Portfolio (optional, default =5) ### Main Program ```shell $ python Init_StockALL_Sp.py $ python stock_index_pro.py $ python main_pro.py ``` ### Output - Daily Trading Data in [Stock Pool](https://github.com/FDUJiaG/QT_Test/blob/master/sqlT_to_csv/stock_all.csv) and [Base Index](https://github.com/FDUJiaG/QT_Test/blob/master/sqlT_to_csv/stock_index_pro.csv) - [Result](https://github.com/FDUJiaG/QT_Test/blob/master/sqlT_to_csv/model_ev_resu.csv) of SVM Model Evaluation - The [Capital Situation](https://github.com/FDUJiaG/QT_Test/blob/master/sqlT_to_csv/my_capital.csv) during Loopback Test - The [Stocks Holding](https://github.com/FDUJiaG/QT_Test/blob/master/sqlT_to_csv/my_stock_pool.csv) in Last Loopback Test Day - [Effect Index](https://github.com/FDUJiaG/QT_Test/blob/master/imag/LoopBack.png) of Quantization - Visualization of Return and Withdrawal ## Dependencies 测试使用的Python版本：3.6.8 测试使用的Anaconda版本：1.9.6 ### Installation or Upgrade for Tushare ```shell $ pip install tushare $ pip install tushare --upgrade ``` ### Import Tushare ```python import tushare as ts ``` tushare版本需大于1.2.10 ### Set Token ```python ts.set_token('your token') ``` 完成调取tushare数据凭证的设置，通常只需要设置一次 ### Initialize Pro API ```python pro = ts.pro_api() # 或者在初始化中直接设置token pro = ts.pro_api('your token') ``` ### Main Data API ```python pro.daily() # 获取日K数据（未赋权） pro.index_daily() # 获取指数行情 pro.trade_cal() # 获取交易日历 ``` ### Package #### Time Handle ```python import datetime ``` #### MySql Handle ```python import pymysql.cursors import sqlalchemy ``` #### Data Handle ```python import numpy as np import pandas as pd from sklearn import svm import pylab as * import math ``` ## 设计过程 ### 数据采集预处理后建模 - 基于[Tushare](https://tushare.pro/document/1?doc_id=131)进行交易数据采集（[股票](https://github.com/FDUJiaG/QT_Test/blob/master/codes/Init_StockALL_Sp.py)，[指数](https://github.com/FDUJiaG/QT_Test/blob/master/codes/stock_index_pro.py)） - 简单[数据预处理](https://github.com/FDUJiaG/QT_Test/blob/master/codes/DC.py)，生成训练集 - 利用[SVM](https://blog.csdn.net/b285795298/article/details/81977271)算法进行[建模](https://github.com/FDUJiaG/QT_Test/blob/master/codes/Model_Evaluate.py)，并[预测涨跌](https://github.com/FDUJiaG/QT_Test/blob/master/codes/SVM.py)情况，准备开发择时策略 ### 模型评估和仓位管理 - 测试区间内[评估指标](https://blog.csdn.net/zhihua_oba/article/details/78677469)的[计算](https://github.com/FDUJiaG/QT_Test/blob/master/codes/Model_Evaluate.py)，包括：Precision，Recall，F1，Negative_Accuracy等值 - 基于[马科维茨理论](https://mp.weixin.qq.com/s/neCSaWK0c4jzWwCfDVFA6A)的[仓位管理](https://github.com/FDUJiaG/QT_Test/blob/master/codes/Portfolio.py)分配，取**次最小的特征值和特征向量**（最佳收益方向） ### 模拟交易测试及回测 - 模拟交易，包括：获取资金账户[数据](https://github.com/FDUJiaG/QT_Test/blob/master/codes/Deal.py)，执行买卖[操作](https://github.com/FDUJiaG/QT_Test/blob/master/codes/Operator.py)，更新持仓天数及买卖[逻辑](https://github.com/FDUJiaG/QT_Test/blob/master/codes/Filter.py)，更新资产表[数据](https://github.com/FDUJiaG/QT_Test/blob/master/codes/Cap_Update_daily.py)等 - 策略框架下，进行[回测](https://github.com/FDUJiaG/QT_Test/blob/master/codes/main_pro.py)并计时 - 计算并返回量化策略[评估指标](https://www.jianshu.com/p/363aa2dd3441)：Return，Withdrawal，Sharp，Risk，IR及Tracking Error等 - 对于Return，Withdrawal的可视化展示 ## [阅读完整版](https://fdujiag.github.io/QT_Test)