将 vars pr 迭代保存到 df，完成后将 df 保存到 csv

Question

我需要结合公司股票代码（Apple Computers 的 AAPL 等）创建一个数据帧 (df_max_res)，其中包含我的股票策略中的 15 个最佳表现。我有一个包含 500 多个股票代码的列表，我使用我自己的四种策略对其进行分析。

在for eachP in perf_array 嵌套内部迭代中，我从策略和代码的所有组合中获得性能结果。我想使用此代码（或更好的建议）将这些结果保存到 DataFrame 和 csv 文件中：

#==============================================================================
#     Saving results in pandas and to a csv-file
#==============================================================================
def saving_res_pandas():
    global df_res, df_max_res
    df_res = pd.DataFrame(columns=('Strategy', 'Ticker', 'Strat', 
                                   'ROI', 'Sharpe R', 'VaR'))
    for eachP in perf_array:
        df_res.loc[len(df_res) + 1] = [strategy, ticker, strat, stratROI]  
    # Select the top 15 of all results (ticker/strategy combo) into new df.
    df_max_res = df_res[:15]           
    # Saving to a csv.
    df_max_res.to_csv('df_performance_data_sp500ish.csv')
    print('After analysing %1.1f Years ~ %d workdays - %d strategies and %d tickers' '\n'
          'The following matrix of tickers and strategies show highest ROI: ' 
          % (years, days, len(strategies), len(stock_list))
         )

    return df_res
#==============================================================================
# Chose which of below methods to save perf-data to disk with
#==============================================================================
saving_res_pandas()

# Reading in df_max_res with best ticker/strategy results
df_max_res = pd.read_csv('df_performance_data_sp500ish.csv')
print(df_max_res)

上面的代码可以很好地创建我的 DataFrame，但它并没有像我预期的那样保存迭代性能结果。

我得到这个输出：

=======================================================
 aa   ===   <function strategy1 at 0x00000000159A0BF8>   ==
=======================================================


Holdings: 0
Funds: 14659

Starting Valuation:  USD 15000.00 ~ DKK: 100000.50
Current Valuation:   USD 14659.05 ~ DKK: 97727.49

===  aa  == <function strategy1 at 0x00000000159A0BF8> ==
ROI: -1.9 perc. & Annual Profit -1894 DKK  ==
######################################################################

cannot set a row with mismatched columns

== ALL Tickers Done for ==  <function strategy1 at 0x00000000159A0BF8> ==================
Strategy analysis pr ticker - COMPLETE !

Empty DataFrame
Columns: [Unnamed: 0, Strategy, Ticker, ROI, SharpeR, VaR]
Index: []

Answer 1

我已尝试减少您的代码以使其更具可读性：

 1. def saving_res_pandas():
 2.     cols = ('Strategy', 'Ticker', 'Strat', 'ROI', 'Sharpe R', 'VaR')
 3.     df_res = pd.DataFrame(columns=cols)
 4.     for _ in perf_array:
 5.         df_res.loc[len(df_res) + 1] = [strategy, ticker, strat, stratROI]  
 6.     # Select the top x of all results (ticker/strategy combo) into new df.
 7.     df_max_res = df_res[:15]           
 8.     df_max_res.to_csv('df_performance_data_sp500ish.csv')
 9.     print('After analysing {0:.1f} Years ~ {1} workdays - {2} strategies and {3} tickers' '\n'
10.           'The following matrix of tickers and strategies show highest ROI: '  
11.           .format(years, days, len(strategies), len(stock_list)))
12.     return df_res

根据上面的代码，我有两个问题：

1. 在第 5 行，strategy, ticker, strat and stratROI 的值是如何获得的

2. 在第 7 行，您正在获取 df_res 的前 15 项，但 DataFrame 尚未排序。在您的原始代码中，下面的排序行已被注释掉（因此我在编辑中将其删除）。

   df_res.reset_index().sort(['ROI', 'VaR', 'Sharpe R'], ascending=[0,1,0])

当您说您想要 15 项最佳表现时，是根据哪个指标（ROI、Var、Sharpe 等）？

Answer 2

我终于想出了解决我烦恼的正确答案。

我是这样解决的：

在 for 循环之前：

# Creating the df that will save my results in the backtest iterations
cols = ('Strategy','Ticker','ROI')  # ,'Sharpe R','VaR','Strat'
df_res = pd.DataFrame(columns = cols)

在 for 和嵌套的 for 循环中

def saving_res_pandas():
    global df_res, df_max_res
    df_res = df_res.append({'Ticker':ticker,'Strategy':strategy, 'ROI':stratROI,}, ignore_index = True)

    return df_res

for 循环的外部和之后：

        df_res = df_res.sort(['ROI'], ascending=[0])
        df_max_res = df_res.head(15)           # Select the top x of all results (ticker/strategy combo) into new df
        # saving to a csv #
        df_max_res.to_csv('df_performance_data_sp500ish.csv')

    print('After analysing %1.1f Years ~ %d workdays - %d strategies and %d tickers' '\n'
    'The following matrix of tickers and strategies show highest ROI:' %(years, days, len(strategies), len(stock_list))
    )
    print()
    print(df_max_res)

感谢您的所有帮助和启发。