Pandas 移动平均计算缺少日期？

Question

I am trying to make a simple moving average using the live forex data from Alpha Vantage (API key can be registered for free from Alpha Vantage). Everything seems to work fine except that the period for the SMA is not complete.

For example, if I set data[:'2020-1-1'] , it should return data from 1.1.2020 up to current date. However, what happens is that the the period from December 2020 to January 2021 is gone.

I tried plotting the graph and realize that the larger my moving average period is, the more recent data are being removed. The plot on my graph is as follows:

GBP USD

Dataframe for beginning period 2020-1-1

Dataframe period fixed but NAN values

Below are my codes in 3 separate files:

This is the execution.py file:

from alpha_vantage.foreignexchange import ForeignExchange
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
from to_USD import currencyExchangeB
from smvgavg import sma

UniSymbol = 'USD'
fromSymbolsB = ['EUR','GBP']

for symbol in fromSymbolsB:
    # store the result that return from "currencyExchange()" function
    result = currencyExchangeB(UniSymbol,symbol)

    # generate graph by passing those result of each currency data
    sma(result,UniSymbol,symbol)

The next one is the to_USD.py file where I use it to pull the foreign currency data from Alpha Vantage:

from alpha_vantage.foreignexchange import ForeignExchange
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
import datetime

# added a class to collect a collection of variable so it can be return easily in the following function
class GRAPH_STRUCT:
    date:any
    symbol:str
    def __init__(self,date,symbol):
        self.date = date
        self.symbol = symbol

def currencyExchangeB(toCurrencyB,fromCurrencyB):
    # create an array to store the result
    result = []

    cc = ForeignExchange(key='%ALPHA_VANTAGE_APIKEY%',output_format='pandas')
    data, meta_data = cc.get_currency_exchange_daily(from_symbol=fromCurrencyB,to_symbol=toCurrencyB,outputsize='full')

    # append those result in the array
    result = GRAPH_STRUCT(data[:'2020-1-1'],toCurrencyB)

    # return those result in the end of the function
    return result

The last file is the smvgavg.py file used to calculate the simple moving average:

import matplotlib.pyplot as plt
import pandas as pd
import numpy as np

#dataObj is a datatype of GRAPH_STRUCT
def sma(dataObj,UniSymbol,symbol):
    maA = 25
    maB = 50
    maT = [maA,maB]
    for ma in maT:
        smaString = "SMA" + str(ma)

        data = dataObj.date
        data[smaString] = data.iloc[:,3].rolling(window = ma).mean()
        data = data.iloc[ma:]

        print(data)
        fig = plt.gcf()
        fig.set_size_inches(10, 6)
        plt.plot(data['4. close'], label='Close' if ma == 0 else "",color='red')
        plt.plot(data['SMA' + str(ma)],label='SMA' + str(ma))

    plt.title(symbol + '/' + UniSymbol, y=1)
    plt.xlabel("Date")
    plt.ylabel("Exchange Rate")
    plt.legend()
    plt.show()

Any help would be appreciated. Thanks.

Answer 1

This isn't because of pandas or Python, rather, it's because of how the SMA is calculated. Unless you do the computation yourself just to see what's going on, it can be kind of tricky.

The formula for the simple moving average is to take n amount of closing prices at n interval. And divide by n . Where P is the closing price and the number is an even interval in your n -day SMA, one period of a 5-day SMA would look something like this:

(P1 + P2 + P3 + P4 + P5)/5

The reason some of your starting dates appear to be missing is because the data point calculated from a period is placed at the end of that period. For example, let's have a look at Airbnb (ABNB) on TD Ameritrade's thinkorswim . This stock was just recently listed, so we can easily see that this 50-day SMA (purple line) also has some dates that appear to be missing:

What you can do is get earlier information that you are not going to display in your graph.

For example, if you want to display 10-day SMA line for the dates December 2020 to February 2021, you could factor in the data 10 days before your start date, and then just graph your start date to the end date. This would fill in that area that appears to be missing because your calculations now account for the previous period (that won't be graphed).