crypto_price_history_gathering_funcs.py

def getPrices(ticker, training_days, tweet_df, mov_avg, forecast_period):
    """
    Using a tweet DataFrame to find the first day and training data from the 
    previous N days to make predictions that coincide with the sentiment date range.
    
    In order to properly backtest with sentiment, a certain amount of training 
    days need to come before the initial date from the tweet DF as well 
    as the moving average and forecast period in order to accurately line up
    with the tweet DF.
    
    Also prepares the price data for FBProphet forecasting.
    """
    # The first day from the tweet DF minus the amount of days to train with, moving avg, and forecast period
    ago = datetime.strptime(
        tweet_df.index[0], 
        "%Y-%m-%d"
    ) - timedelta(
        days=training_days+mov_avg+forecast_period
    )
    
    # The last day of the tweet DF
    now = datetime.strptime(
        tweet_df.index[-1], 
        "%Y-%m-%d"
    )
    
    # Getting prices
    prices = pd.DataFrame(
        client.get_prices_eod(
            ticker+"-USD.CC", 
            from_=ago.strftime("%Y-%m-%d"),
            to=now.strftime("%Y-%m-%d")
        )
    )

    # Set index
    prices = prices.set_index('date', drop=True)
    
    # Getting the N Day Moving Average and rounding the values for some light data preprocessing
    prices['MA'] = prices[['open']].rolling(
        window=mov_avg
    ).mean().apply(lambda x: round(x, 2))

    # Dropping Nans
    prices.dropna(inplace=True)

    # Resetting format for FBP
    prices = prices.reset_index().rename(
        columns={"date": "ds", "MA": "y"}
    )
    
    return prices