Анализ корреляций биржевых колебаний

# !conda install snappy python-snappy pyarrow fastparquet 
# !pip install pandas_datareader

%matplotlib inline 
%config InlineBackend.figure_format = 'svg' #векторный формат

from matplotlib import pyplot as plt

plt.style.use('ggplot')  # Красивые графики
plt.rcParams['figure.figsize'] = (12, 5)  # Размер картинок

import numpy as np
import pandas as pd
from datetime import timedelta
from datetime import datetime
import math
import os
import pandas_datareader.data as web

tikers = 'SBER,AFKS,GAZP,MAIL,YNDX,FXUS,FXIT,FXCN,FXDE,FXWO,FXRW'.split(',')

def download_data(tiker, path):
    filepath = os.path.join(path, f'{tiker}_2020.parquet')
    if os.path.exists(filepath):
        return
    df = web.DataReader(tiker, 'moex', start='2020-01-01', end='2020-11-12')
    df.to_parquet(filepath, engine='fastparquet')

for tiker in tikers:
    print(f'downloading: {tiker}')
    download_data(tiker, "D:\\data2\\Documents\\notebooks")

def load_data(tiker, path):
    result = None
    filepath = os.path.join(path, f'{tiker}_2020.parquet')
    if os.path.exists(filepath):
        result = pd.read_parquet(filepath)
    return result

dfs = {}
for tiker in tikers:
    dfs[tiker] = load_data(tiker, "D:\\data2\\Documents\\notebooks")

dfs['FXIT']['BOARDID'].unique(), dfs['YNDX']['BOARDID'].unique()

def price_data(x_df, boardid='TQBR'):   
    xdf = x_df.copy()
    xdf = xdf.where(xdf['BOARDID'] == boardid)
    xdf.dropna(axis=0, how='all', inplace=True)
    xdf.dropna(axis=1, how='all', inplace=True)
    if 0 == xdf.size:
        return xdf
    xdf['OPEN']    = xdf['OPEN'].interpolate(method='polynomial', order=2);
    xdf['CLOSE']   = xdf['CLOSE'].interpolate(method='polynomial', order=2);
    xdf['WAPRICE'] = xdf['WAPRICE'].interpolate(method='polynomial', order=2);
    return xdf

# для простоты, считаем что если имя начинается на FX то это ETF
for tiker in tikers:
    dfs[tiker] = price_data(dfs[tiker], boardid='TQTF' if 'FX' in tiker else 'TQBR')

# WAPRICE/median
def calc_wdm(df):
    return df['WAPRICE'] / df['WAPRICE'].median()

for tiker in tikers:
    dfs[tiker][tiker] = calc_wdm(dfs[tiker])

df_base = dfs['FXIT'][[]]

def merge_dfs(base):
    result = base
    for tiker in tikers:
        result = result.merge(dfs[tiker][[tiker]], how='outer', on='TRADEDATE')
    return result
        

xxdf = merge_dfs(df_base)

xxdf['diff_us'] = xxdf['FXUS'] - xxdf['FXIT']

# рисуем графики
xxdf.plot(y='FXUS,FXIT,FXCN,FXDE,FXWO,FXRW'.split(','), kind='line', title='ETF: WAPRICE / median')
xxdf.plot(y='SBER,AFKS,GAZP,MAIL,YNDX'.split(','), kind='line', title='RU: WAPRICE / median')
xxdf.plot(y='diff_us'.split(','), kind='line', title='FXUS - FXIT')
xxdf.plot(y='FXIT,YNDX'.split(','), kind='line', title='FXIT & YNDX')
xxdf.plot(y='FXWO,FXRW,FXUS'.split(','), kind='line', title='FXUS & FXWO & FXRW');