Submission

YuAn_Lin/README.md

@@ -0,0 +1,10 @@
+<!-- ABOUT THE PROJECT -->
+## About The Project
+
+There are six main files in this project.
+  * data.py: Contain functions for downloading universe of stocks (S&P500 constituents).
+  * series_generator.py: Contain functions for generating returns series described in the paper.
+  * stop_loss.py: Contain a class for constructing a framework to determine if adding a stop-loss policy would improve the results of the strategy.
+  * Demo.ipynb: Jupyter Notebook for demo.
+  * Demo.html: HTML for demonstrate visualization using Bokeh.
+  * unit_test.py: Unit tests

YuAn_Lin/data.py

@@ -0,0 +1,60 @@
+#For crawling S&P500 Tickers
+import os
+import requests
+from io import StringIO
+
+#For getting and modifying data
+import numpy as np
+import pandas as pd
+import datetime
+import yfinance as yf
+
+
+def get_sp500_tickers():
+    """Crawling S&P500 Tickers
+
+    Returns
+    -------
+    List contains S&P500 tickers
+    """
+
+    # Send a request to the link
+    headers = { 'user-agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/84.0.4147.105 Safari/537.36'}
+    link = "https://www.slickcharts.com/sp500"
+    res = requests.get(link, headers = headers)
+
+    # Clean the text, and save as a dataframe 
+    lines = res.text.replace("\r", "").split("\n")
+    df = pd.read_html( StringIO("\n".join( lines[:] ) ), header = None )[0]
+    tickers = df["Symbol"]
+    tickers = tickers.apply(lambda s: s.replace(".", "-"))  # Modify tickers, so they can fit Yahoo Finance's ticker format
+
+    return list(tickers.values)
+
+def get_adj_close(tickers, start_y, start_m, start_d, end_y, end_m, end_d):
+    """Get adjusted close data from Yahoo Finance
+
+    Parameters
+    ----------
+    tickers : str, list of str
+        List of tickers to download
+    start_y, start_m, start_d : int
+        start year, start month, start day
+    end_y, end_m, end_d : int
+        end year, end month, end day
+
+    Returns
+    -------
+    Pandas DataFrame contains adjusted close
+    """
+
+    # Modify date format
+    start = datetime.datetime(start_y, start_m, start_d).strftime("%Y-%m-%d")
+    end = datetime.datetime(end_y, end_m, end_d).strftime("%Y-%m-%d")
+
+    # Get data using yfinance module
+    data = yf.download(tickers, start, end)
+    data = data['Adj Close']
+    data = data.round(decimals=2)
+
+    return data

YuAn_Lin/series_generator.py

@@ -0,0 +1,285 @@
+import pandas as pd
+import numpy as np
+import datetime
+
+import matplotlib.pyplot as plt
+from bokeh.io import output_notebook, show
+from bokeh.plotting import figure
+from bokeh.models import CheckboxGroup, CustomJS, ColumnDataSource, Panel, Tabs, ColorBar, LinearColorMapper
+from bokeh.transform import transform
+from bokeh.layouts import row, column
+from bokeh.palettes import Spectral10, Reds
+from bokeh.models.widgets import DataTable, DateFormatter, TableColumn
+
+
+def generate_date(start_date: str, end_date: str, step:int = 1, ignore_weekends: bool = True):
+    """Generate list of datetimes
+
+    Parameters
+    ----------
+    start_date : str in %Y%m%d format
+        Start date
+
+    end_date : str in %Y%m%d format
+        End date
+
+    step : int
+        Interval between two dates
+
+    ignore_weekends : bool, default True
+        Whether the process ignores weekends.
+
+    Returns
+    -------
+    List of datetimes
+    """
+
+    # Generate sequence of dates
+    start_date = datetime.datetime.strptime(start_date, '%Y%m%d').date()
+    end_date = datetime.datetime.strptime(end_date, '%Y%m%d').date()
+    number_of_days = ((end_date - start_date).days + 1)
+    dates = [start_date + datetime.timedelta(days = i) for i in range(number_of_days)]
+
+    if ignore_weekends:
+        dates = [d for d in dates if not d.isoweekday() in [6,7]]
+
+    dates = [dates[i] for i in range(len(dates)) if i%step == 0]
+
+    return dates
+
+
+def random_walk_generater(mean: float, std: float, start_date: str, end_date: str, step:int = 1, ignore_weekends: bool = True):
+    """Generate Series contains returns through Random Walk Hypothesis
+
+    Method:
+    rt = µ + et, et ∼ White Noise(0, σ^2),
+
+    Parameters
+    ----------
+    mean : float
+        Mean of the process
+
+    std : float
+        Standard deviation of the white noise
+
+    start_date : str in %Y%m%d format
+        Start date of the process
+
+    end_date : str in %Y%m%d format
+        End date of the process
+
+    step : int
+        Interval between two dates
+
+    ignore_weekends : bool, default True
+        Whether the process ignores weekends.
+
+    Returns
+    -------
+    Series contains returns
+    """
+
+    # Generate sequence of dates
+    dates = generate_date(start_date, end_date, step, ignore_weekends)
+
+    # Generate white noise
+    white_noise = np.random.normal(0, std, size = len(dates))
+
+    # Generate returns Series
+    rt = pd.Series(mean + white_noise, index = dates, name = 'random_walk')
+    rt.index = pd.to_datetime(rt.index)
+    rt.index.name = "Date"
+
+    return rt
+
+
+def ar_1_return_generater(mean: float, std: float, rho: float, r0: float, start_date: str, end_date: str, step:int = 1, ignore_weekends: bool = True):
+    """Generate Series contains returns through the process described in Kaminski, Kathryn M., and Andrew W. Lo. (2013) equation 14.
+
+    Method:
+    rt = µ + ρ(rt-1 - µ) + et, et ∼ White Noise(0, σ^2),
+
+    Parameters
+    ----------
+    mean : float
+        Mean of the process
+
+    std : float
+        Standard deviation of the white noise
+
+    rho : float
+        ρ ∈ (−1, 1)
+
+    r0 : float
+        Initial value of the series
+
+    start_date : str in %Y%m%d format
+        Start date of the process
+
+    end_date : str in %Y%m%d format
+        End date of the process
+
+    step : int
+        Interval between two dates
+
+    ignore_weekends : bool, default True
+        Whether the process ignores weekends.
+
+    Returns
+    -------
+    Series contains returns
+    """
+
+    # Generate sequence of dates
+    dates = generate_date(start_date, end_date, step, ignore_weekends)
+
+    # Generate white noise
+    white_noise = np.random.normal(0, std, size = len(dates))
+
+    # Generate returns Series
+    rt = pd.Series(r0, index = dates, name = 'ar_1')
+    for i in range(1, len(rt)):
+        rt.iloc[i] = mean + rho*(rt.iloc[i - 1] - mean) + white_noise[i]
+
+    rt.index = pd.to_datetime(rt.index)
+    rt.index.name = "Date"
+
+    return rt
+
+
+def regime_switching_return_generater(mean_1: float, std_1: float, mean_2: float, std_2: float, I0: float, trans_prob_matrix: np.array, 
+                                      start_date: str, end_date: str, step:int = 1, ignore_weekends: bool = True):
+    """Generate Series contains returns through the process described in Kaminski, Kathryn M., and Andrew W. Lo. (2013) equation 19.
+
+    Method:
+    rt = It*r1t + (1 − It)*r2t, rit ~ N(µi, σi^2)
+
+                It+1 = 1     It+1 = 0
+    A ≡ It = 1  [p11          p12]
+
+        It = 0  [p21          p22]
+
+    where A is the Markov transition probabilities matrix that governs the transitions between the two states.
+
+    Parameters
+    ----------
+    mean_1 : float
+        Mean of r1
+
+    std_1 : float
+        Standard deviation of r1
+
+    mean_2 : float
+        Mean of r2
+
+    std_2 : float
+        Standard deviation of r2
+
+    I0 : float
+        Initial value of the state
+
+    trans_prob_matrix : 2x2 np.array
+        Markov transition probabilities matrix
+
+    start_date : str in %Y%m%d format
+        Start date of the process
+
+    end_date : str in %Y%m%d format
+        End date of the process
+
+    step : int
+        Interval between two dates
+
+    ignore_weekends : bool, default True
+        Whether the process ignores weekends.
+
+    Returns
+    -------
+    1. Series contains returns
+    2. Series contains states
+    """
+
+    # Generate sequence of dates
+    dates = generate_date(start_date, end_date, step, ignore_weekends)
+
+    # Generate r1, r2
+    r1 = np.random.normal(mean_1, std_1, size = len(dates))
+    r1 = pd.Series(r1, index = dates)
+
+    r2 = np.random.normal(mean_2, std_2, size = len(dates))
+    r2 = pd.Series(r2, index = dates)
+
+    # Generate returns Series
+    It = pd.Series(I0, index = dates)
+    for i in range(1, len(It)):
+        if It.iloc[i - 1] == 1:
+            It.iloc[i] = np.random.choice([1, 0], 1, p = trans_prob_matrix[0])[0]
+        else:
+            It.iloc[i] = np.random.choice([1, 0], 1, p = trans_prob_matrix[1])[0]
+
+    rt = It*r1 + (1 - It)*r2
+    rt.index = pd.to_datetime(rt.index)
+    rt.index.name = "Date"
+    rt.name = 'regime_switching'
+
+    return rt, It
+
+
+def trend_chart(returns_series, compounding: bool = False, height: int = 350, width: int = 800):
+    """Trend chart of the result using Bokeh.
+
+    Parameters
+    ----------
+    returns_series : Pandas Series
+        Series contains returns.
+
+    compounding : bool, default False
+        Whether returns are reinvested back into the account.
+
+    height : int
+        Height of the plot
+
+    width : int
+        Width of the plot
+
+    Returns
+    -------
+    None
+    """
+
+    data = returns_series
+
+    if compounding:
+        cum = (data + 1).cumprod()
+    else:
+        cum = data.cumsum() + 1
+
+    cum = pd.DataFrame(cum)
+    source = ColumnDataSource(data = cum)
+
+    p = figure(x_axis_type="datetime", title="Trend Line", plot_height=height, plot_width=width)
+    p.xgrid.grid_line_color=None
+    p.ygrid.grid_line_alpha=0.5
+    p.xaxis.axis_label = 'Time'
+    p.yaxis.axis_label = 'Total Return'
+    lines = []
+    for i in range(len(cum.columns)):
+        lines.append(p.line("Date", cum.columns[i], source=source, line_width=2, line_alpha=0.8, line_color = Spectral10[i%10], legend_label = cum.columns[i], muted_color = Spectral10[i%10], muted_alpha = 0.1))
+
+    p.legend.location = "top_left"
+    p.legend.click_policy="mute"
+
+    LABELS = list(cum.columns)
+    checkbox_group = CheckboxGroup(labels=LABELS)
+    checkbox_group.active = list(range(len(LABELS)))
+
+    code = """ for (var i = 0; i < lines.length; i++) {
+                    lines[i].visible = false;
+                    if (cb_obj.active.includes(i)){lines[i].visible = true;}
+                }
+           """
+    callback = CustomJS(code = code, args = {'lines': lines})
+    checkbox_group.js_on_click(callback)
+
+    layout = row(p, checkbox_group)
+    show(layout)