Getting returns merged with fundamental dataframe

Here is a quick example algo I coded up, that demonstrate the problem. Furthermore, I am also getting an trying to dynamically access a method that does not exist error that referes to QuantConnect.Data.UniverseSelection.SecurityChanges, but I do not know how

Here is the code:

from AlgorithmImports import *
from datetime import timedelta
import pandas as pd
# endregion

class CalmTanLlama(QCAlgorithm):

    def Initialize(self):
        self.SetStartDate(2020, 10, 15)  # Set Start Date
        self.SetEndDate(2020, 11, 20)
        self.SetCash(100000)  # Set Strategy Cash
        self.SetBenchmark("SPY")
        self.UniverseSettings.Resolution = Resolution.Hour
        self.SetBrokerageModel(BrokerageName.InteractiveBrokersBrokerage, AccountType.Margin)
        
        self.numCoarse = 200
        self.numFineMarket = 10
        self.numFine = 10
        self.rebalancingFunc = Expiry.EndOfMonth
        self.returnLookback = timedelta(weeks = 104)
        self.returnResolution = Resolution.Daily
        self.nextRebalance = None

        self.AddUniverse(self.CoarseUniverse, self.SelectFine)
        self.SetAlpha(ConstantAlphaModel(InsightType.Price, InsightDirection.Up, timedelta(hours=5)))
        self.SetPortfolioConstruction(NullPortfolioConstructionModel())
        self.SetExecution(NullExecutionModel())
        self.SetRiskManagement(NullRiskManagementModel)

    def CoarseUniverse(self, coarse):
        if self.nextRebalance is not None and self.Time < self.nextRebalance:
            return Universe.Unchanged
        self.nextRebalance = self.rebalancingFunc(self.Time)
        
        selected = sorted([x for x in coarse if x.HasFundamentalData and x.Price > 5], 
                            key = lambda x: x.DollarVolume, reverse=True)
        
        return [x.Symbol for x in selected[:self.numCoarse]]

    def SelectFine(self, fine):
        
        sortedMarketCap = sorted(fine, key = lambda x: x.MarketCap, reverse=True)[:self.numFineMarket]

        #Get the results of the linear regression analysis
        linearRegressionResults = self.AppendDataFrames(self, sortedMarketCap, self.returnLookback, self.numFine)
        ##
        #Do something with the dataframe
        ##


        return [x.Symbol for x in sortedMarketCap]



    def AppendDataFrames(self, algorithm, symbols, returnLookback, maxstocks):
    
        securities = [x for x in symbols]
        
        #our fundamental values
        fundamentals = ["ValuationRatios.SustainableGrowthRate", "ValuationRatios.PayoutRatio", "ValuationRatios.CashReturn", "ValuationRatios.FCFRatio", "ValuationRatios.EarningYield", "ValuationRatios.SalesYield", "ValuationRatios.BookValueYield", "ValuationRatios.PBRatio",
                "ValuationRatios.PEGRatio", "ValuationRatios.BuyBackYield", "ValuationRatios.PricetoEBITDA", "ValuationRatios.ForwardPERatio", "ValuationRatios.EVToEBITDA3YrAvgChange",
                "ValuationRatios.EVtoFCF", "ValuationRatios.EVtoEBIT", "ValuationRatios.EVToForwardEBITDA", "ValuationRatios.EVToForwardRevenue", "ValuationRatios.ForwardROE",
                "ValuationRatios.ForwardDividend"]

        #setting empty dataframe that we can concat with
        dataframe = pd.DataFrame()

        #Get the returns of the stocks, and get them shaped proberly
        dataframe = algorithm.History([x.Symbol for x in securities], returnLookback, Resolution.Daily)
        dataframe = dataframe.drop_duplicates().close.unstack(level=0)
        dataframe = dataframe.iloc[0] / dataframe.iloc[-1]
        dataframe = pd.DataFrame(dataframe)
        dataframe = dataframe.rename(columns={0: 'returns'})

        fundamental_dataframe = pd.DataFrame()

        #we need to get the symbols and symbol object, so we can set the name in the dataframe, and also get the fundamental value (ticker = setting name)
        for fundamental in securities:
            ratio =   [fundamental.ValuationRatios.SustainableGrowthRate,
                            fundamental.ValuationRatios.PayoutRatio,
                            fundamental.ValuationRatios.CashReturn,
                            fundamental.ValuationRatios.FCFRatio,
                            fundamental.ValuationRatios.EarningYield,
                            fundamental.ValuationRatios.SalesYield,
                            fundamental.ValuationRatios.BookValueYield,
                            fundamental.ValuationRatios.PEGRatio,
                            fundamental.ValuationRatios.PBRatio,
                            fundamental.ValuationRatios.BuyBackYield,
                            fundamental.ValuationRatios.PricetoEBITDA,
                            fundamental.ValuationRatios.ForwardPERatio,
                            fundamental.ValuationRatios.EVToEBITDA3YrAvgChange,
                            fundamental.ValuationRatios.EVtoFCF,
                            fundamental.ValuationRatios.EVtoEBIT,
                            fundamental.ValuationRatios.EVToForwardEBITDA,
                            fundamental.ValuationRatios.EVToForwardRevenue,
                            fundamental.ValuationRatios.ForwardROE,
                            fundamental.ValuationRatios.ForwardDividend]

            #transpose the dataframe, so it is a long dataframe like in the notebook, set the column names, and concat
            ratio = pd.DataFrame(ratio, index = fundamentals, columns=[fundamental.Symbol])
            ratio = ratio.transpose()
            fundamental_dataframe = pd.concat([ratio, fundamental_dataframe], axis=0)
                            
        #Make the dataframe into a nicely shaped dataframe


        #Append the 2 dataframes
        dataframe = pd.concat([fundamental_dataframe, dataframe], axis=1)

        #
        #Do something with the dataframe
        #
        return 0

Have a good day

Lucas

Hi Lucas,

In order to create a dataframe combining the fundamentals and history data, we'll need

1. Use self.Securities[symbol].Fundamentals instead of the Security object itself.
2. Pass a universe with Fine Fundamental Filter to warm up the fundamentals.
3. Concatenate them with the same name to avoid duplicate index with NaNs

Please find the attached backtest as an example.

Best
Louis

Hi Louis

I appriciate the help a lot. I will port this into the alpha module, seems straightforward. 

A little follow-up question if you dont mind. Why does this method not work in the FineUniverse? 

Have a good day

Lucas

Hi Lucas,
We can use the same inside the fine filter. We need to use the fine fundamental object to get the fundamentals.
Best,
Varad Kabade