Help with CVXPY for Mean Variance Optimization

Hey, would you be able to provide your code so we can see what you're trying to achieve? Seeing what's inside your scheduled event function (MonthStart?) at least would be helpful.

Looks like you're trying to access a non-existing element from a list or something. This usually happens when you try to access something from the incoming data slice - you should always check first if the element you're looking for exists before using it.

Here is the code :

from math import ceil,floor,isnan
from datetime import datetime
import pandas as pd
import numpy as np
import cvxpy as cv

class AssetAllocationAlgorithm(QCAlgorithm):
    
    def Initialize(self):
        self.SetStartDate(2016, 1, 1)  #Set Start Date
        self.SetEndDate(2018, 1, 1)    #Set End Date
        self.SetCash(100000)            #Set Strategy Cash

        tickers = ["IEF", "TLT", "SPY", "EFA", "EEM", "JPXN", "XLK"]
        self.symbols = [] 
        for i in tickers:
            self.symbols.append(self.AddEquity(i, Resolution.Daily).Symbol)
        for syl in self.symbols:
            syl.window = RollingWindow[TradeBar](252) 

        self.Schedule.On(self.DateRules.MonthStart("SPY"), self.TimeRules.AfterMarketOpen("SPY"), Action(self.Rebalancing))
       

    def OnData(self, data):
        if data.ContainsKey("SPY"):
            for syl in self.symbols:
                syl.window.Add(data[syl])

    def Rebalancing(self):
        data = {}
        for syl in self.symbols:
            data[syl] = [float(i.Close) for i in syl.window]
        df_price = pd.DataFrame(data,columns=data.keys())
        # To order prices from oldest to latest
        df_price = df_price.sort_values(by = 0, ascending = False)
        # Calculating returns
        daily_return = df_price.pct_change().dropna().values

        if daily_return.empty == False:
            a = PortfolioOptimization(daily_return, 0, len(data))
            opt_weight = a.opt_portfolio()  
            
            if isnan(sum(opt_weight)): return
            self.Log(str(opt_weight))
            
            for i in range(len(data)):
                self.SetHoldings(df_price.columns[i], opt_weight[i])
                
                # equally weighted
                # self.SetHoldings(self.symbols[i], 1.0/len(data))

        
class PortfolioOptimization(object):

    import numpy as np
    import pandas as pd
    import cvxpy as cv

    def __init__(self, df_return, risk_free_rate, num_assets):
        
        self.daily_return = df_return
        self.risk_free_rate = risk_free_rate
        self.n = num_assets # numbers of risk assets in portfolio
        self.mu = np.matrix(df_return.mean(axis = 0).values)
        self.sigma = df_return.cov().values

    def opt_portfolio(self):
        
        w = cv.Variable((self.mu.shape[1],1))
        """
        MV model Variables
        """
        risk = cv.quad_form(w, self.sigma) 
        
        constraints = [cv.sum(w) == 1]
        constraints.append(w >= 0)
        constraints.append(w <= 0.25)

        prob = cv.Problem(cv.Minimize(risk), constraints)
        prob.solve()
        
        weights = np.matrix(w.value).T
        weights = weights/np.sum(weights)
        weights = np.squeeze(np.array(weights))

        return weights

Without a line number it's hard to tell where the problem is happening. Given the key that errors is 0, I suspect it may be the following line:

self.SetHoldings(df_price.columns[i], opt_weight[i])

Are you sure df_price.columns[0] and opt_weight[0] exist?

You're going to have to do some troubleshooting in each of the places where this error could occur - i.e. where you're accessing a collection by an ID beginning with 0.

Hi Dany,

I believe your PortfolioOptimization class works fine, and there are two problems with your Rebalancing method:

1. By line df_price = df_price.sort_values(by = 0, ascending = False) , I believe you want to basically reverse all rows. However, the label or index are not values and I don’t think using sort_values() is a good idea here. .iloc[::-1] works just fine here. You can find more information here and here.

2. daily_return = df_price.pct_change().dropna().values gives error: “‘numpy.ndarray’ object has no attribute ‘values’“. I believe you want the 2-D table, and .values should not be necessary here.

Here is an attempt with those modifications. Hope if helps.