Spaces:
Sleeping
Sleeping
import pandas as pd | |
import numpy as np | |
import yfinance as yf | |
import streamlit as st | |
import plotly.graph_objects as go | |
import time | |
import sys | |
with open(r"style/style.css") as css: | |
st.markdown(f"<style>{css.read()}</style>", unsafe_allow_html=True) | |
st.markdown( | |
"<h1 style='text-align: center;'><u>CapiPort</u></h1>", unsafe_allow_html=True | |
) | |
st.markdown( | |
"<h5 style='text-align: center; color: gray;'>Your Portfolio Optimisation Tool</h5>", | |
unsafe_allow_html=True, | |
) | |
st.header( | |
"", | |
divider="rainbow", | |
) | |
color = "Quest" | |
st.markdown( | |
"<h1 style='text-align: center;'>🔍 Quest for financial excellence begins with meticulous portfolio optimization</u></h1>", | |
unsafe_allow_html=True, | |
) | |
st.header( | |
"", | |
divider="rainbow", | |
) | |
list_df = pd.read_csv("Data/Company List.csv") | |
company_name = list_df["Name"].to_list() | |
company_symbol = (list_df["Ticker"] + ".NS").to_list() | |
company_dict = dict() | |
company_symbol_dict = dict() | |
for CSymbol, CName in zip(company_symbol, company_name): | |
company_dict[CName] = CSymbol | |
for CSymbol, CName in zip(company_symbol, company_name): | |
company_symbol_dict[CSymbol] = CName | |
st.markdown( | |
""" | |
<style> | |
.big-font { | |
font-size:20px; | |
} | |
</style>""", | |
unsafe_allow_html=True, | |
) | |
st.markdown('<p class="big-font">Select Multiple Companies</p>', unsafe_allow_html=True) | |
com_sel_name = st.multiselect("", company_name, default=None) | |
com_sel_date = [] | |
for i in com_sel_name: | |
d = st.date_input( | |
f"Select your vacation for next year - {i}", | |
value= pd.Timestamp('2021-01-01'), | |
format="YYYY-MM-DD", | |
) | |
com_sel_date.append(d) | |
com_sel = [company_dict[i] for i in com_sel_name] | |
num_tick = len(com_sel) | |
if num_tick > 1: | |
com_data = pd.DataFrame() | |
for cname, cdate in zip(com_sel, com_sel_date): | |
stock_data_temp = yf.download(cname, start=cdate, end=pd.Timestamp.now().strftime('%Y-%m-%d'))['Adj Close'] | |
stock_data_temp.name = cname | |
com_data = pd.merge(com_data, stock_data_temp, how="outer", right_index=True, left_index=True) | |
for i in com_data.columns: | |
com_data.dropna(axis=1, how='all', inplace=True) | |
# com_data.dropna(inplace=True) | |
num_tick = len(com_data.columns) | |
if num_tick > 1: | |
com_sel_name_temp = [] | |
for i in com_data.columns: | |
com_sel_name_temp.append(company_symbol_dict[i]) | |
com_sel = com_data.columns.to_list() | |
st.dataframe(com_data, use_container_width=True) | |
## Log-Return of Company Dataset | |
log_return = np.log(1 + com_data.pct_change()) | |
## Generate Random Weights | |
rand_weig = np.array(np.random.random(num_tick)) | |
## Rebalancing Random Weights | |
rebal_weig = rand_weig / np.sum(rand_weig) | |
## Calculate the Expected Returns, Annualize it by * 252.0 | |
exp_ret = np.sum((log_return.mean() * rebal_weig) * 252) | |
## Calculate the Expected Volatility, Annualize it by * 252.0 | |
exp_vol = np.sqrt(np.dot(rebal_weig.T, np.dot(log_return.cov() * 252, rebal_weig))) | |
## Calculate the Sharpe Ratio. | |
sharpe_ratio = exp_ret / exp_vol | |
# Put the weights into a data frame to see them better. | |
weights_df = pd.DataFrame( | |
data={ | |
"company_name": com_sel_name_temp, | |
"random_weights": rand_weig, | |
"rebalance_weights": rebal_weig, | |
} | |
) | |
st.divider() | |
st.markdown( | |
"<h5 style='text-align: center;'>Random Portfolio Weights</h5>", | |
unsafe_allow_html=True, | |
) | |
st.dataframe(weights_df, use_container_width=True) | |
# Do the same with the other metrics. | |
metrics_df = pd.DataFrame( | |
data={ | |
"Expected Portfolio Returns": exp_ret, | |
"Expected Portfolio Volatility": exp_vol, | |
"Portfolio Sharpe Ratio": sharpe_ratio, | |
}, | |
index=[0], | |
) | |
st.markdown( | |
"<h5 style='text-align: center;'>Random Weights Metrics</h5>", | |
unsafe_allow_html=True, | |
) | |
st.dataframe(metrics_df, use_container_width=True) | |
st.divider() | |
## Let's get started with Monte Carlo Simulations | |
## How many times should we run Monte Carlo | |
num_of_port = 8000 | |
## Create an Array to store the weights as they are generated | |
all_weights = np.zeros((num_of_port, num_tick)) | |
## Create an Array to store the returns as they are generated | |
ret_arr = np.zeros(num_of_port) | |
## Create an Array to store the volatilities as they are generated | |
vol_arr = np.zeros(num_of_port) | |
## Create an Array to store the Sharpe Ratios as they are generated | |
sharpe_arr = np.zeros(num_of_port) | |
## Track Progress with a Bar | |
progress_text = "Operation in progress. Please wait." | |
my_bar = st.progress(0, text=progress_text) | |
## Let's start the Monte Carlo Simulation | |
for ind in range(num_of_port): # Corrected the range to iterate from 0 to num_of_port | |
time.sleep(0.001) | |
## Let's first Calculate the Weights | |
weig = np.array(np.random.random(num_tick)) | |
weig = weig / np.sum(weig) | |
## Append the Weights to Weigths array | |
all_weights[ind, :] = weig | |
## Calculate and Append the Expected Log Returns to Returns Array | |
ret_arr[ind] = np.sum((log_return.mean() * weig) * 247) | |
## Calculate and Append the Volatility to the Volatitlity Array | |
vol_arr[ind] = np.sqrt(np.dot(weig.T, np.dot(log_return.cov() * 247, weig))) | |
## Calculate and Append the Sharpe Ratio to Sharpe Ratio Array | |
sharpe_arr[ind] = ret_arr[ind] / vol_arr[ind] | |
if ind % 100 == 0: | |
my_bar.progress((ind + 1) / num_of_port, text=progress_text) | |
# clear progress bar | |
my_bar.empty() | |
## Let's create a Data Frame with Weights, Returns, Volatitlity, and the Sharpe Ratio | |
sim_data = [ret_arr, vol_arr, sharpe_arr, all_weights] | |
## Create a Data Frame using above, then Transpose it | |
sim_df = pd.DataFrame(data=sim_data).T | |
## Give the columns in Simulation Data Proper Names | |
sim_df.columns = ["Returns", "Volatility", "Sharpe Ratio", "Portfolio Weights"] | |
## Make sure the Data Types are correct in the Data Frame | |
sim_df = sim_df.infer_objects() | |
# Print out the results. | |
st.write("\n\n") | |
st.markdown( | |
"<h4 style='text-align: center;'>Simulation Results</h4>", | |
unsafe_allow_html=True, | |
) | |
st.dataframe(sim_df.head(), use_container_width=True) | |
# Return the Max Sharpe Ratio from the run. | |
max_sharpe_ratio = sim_df.loc[sim_df["Sharpe Ratio"].idxmax()] | |
# Return the Min Volatility from the run. | |
min_volatility = sim_df.loc[sim_df["Volatility"].idxmin()] | |
max_sharpe_weights_df = pd.DataFrame( | |
data={ | |
"company_name": com_sel_name_temp, | |
"random_weights": max_sharpe_ratio["Portfolio Weights"], | |
} | |
) | |
st.markdown( | |
"<h5 style='text-align: center;'>Portfolio with Max Sharpe Ratio</h5>", | |
unsafe_allow_html=True, | |
) | |
st.dataframe(max_sharpe_ratio, use_container_width=True) | |
st.dataframe(max_sharpe_weights_df, use_container_width=True) | |
min_volatility_weights_df = pd.DataFrame( | |
data={ | |
"company_name": com_sel_name_temp, | |
"random_weights": min_volatility["Portfolio Weights"], | |
} | |
) | |
st.markdown( | |
"<h5 style='text-align: center;'>Portfolio with Min Volatility</h5>", | |
unsafe_allow_html=True, | |
) | |
st.dataframe(min_volatility, use_container_width=True) | |
st.dataframe(min_volatility_weights_df, use_container_width=True) | |
st.divider() | |
st.markdown("<h1 style='text-align: center;'>Plotting</h1>", unsafe_allow_html=True) | |
fig = go.Figure( | |
data=go.Scatter( | |
x=sim_df["Volatility"], | |
y=sim_df["Returns"], | |
mode="markers", | |
marker=dict(color=sim_df["Sharpe Ratio"], colorscale="RdYlBu", size=10), | |
) | |
) | |
# Add color bar | |
fig.update_layout(coloraxis_colorbar=dict(title="Sharpe Ratio")) | |
# Add title and axis labels | |
fig.update_layout( | |
title="Portfolio Returns Vs. Risk", | |
xaxis=dict(title="Standard Deviation / Volatility"), | |
yaxis=dict(title="Returns"), | |
) | |
# Plot the Max Sharpe Ratio, using a `Red Star`. | |
fig.add_trace( | |
go.Scatter( | |
x=[max_sharpe_ratio[1]], | |
y=[max_sharpe_ratio[0]], | |
mode="markers", | |
marker=dict(color="red", symbol="star", size=20), | |
name="Max Sharpe Ratio", | |
) | |
) | |
# Plot the Min Volatility, using a `Blue Star`. | |
fig.add_trace( | |
go.Scatter( | |
x=[min_volatility[1]], | |
y=[min_volatility[0]], | |
mode="markers", | |
marker=dict(color="blue", symbol="star", size=20), | |
name="Min Volatility", | |
) | |
) | |
st.plotly_chart(fig, use_container_width=True) | |