def matshow_tseries(time_series,
fig=None,
axis=0,
xtick_n=5,
time_unit=None,
xlabel=None,
ylabel=None):
"""Creates an image of the time-series, ordered according to the first
dimension of the time-series object
Parameters
----------
time_series: a nitime time-series object
fig: a figure handle, opens a new figure if None
axis: an axis number (if there are several in the figure to be opened),
defaults to 0.
xtick_n: int, optional, sets the number of ticks to be placed on the x axis
"""
if fig is None:
fig = plt.figure()
if not fig.get_axes():
ax = fig.add_subplot(1, 1, 1)
else:
ax = fig.get_axes()[axis]
#Make sure that time displays on the x axis with the units you want:
#If you want to change the time-unit on the visualization from that used to
#represent the time-series:
if time_unit is not None:
tu = time_unit
conv_fac = ts.time_unit_conversion[time_unit]
#Otherwise, get the information from your input:
else:
tu = time_series.time_unit
conv_fac = time_series.time._conversion_factor
this_time = time_series.time / float(conv_fac)
ax.matshow(time_series.data)
ax.set_xticks(list(range(len(this_time)))[::len(this_time) / xtick_n])
ax.set_xticklabels(this_time[::len(this_time) / xtick_n])
if xlabel is None:
ax.set_xlabel('Time (%s)' % tu)
else:
ax.set_xlabel(xlabel)
if ylabel is not None:
ax.set_ylabel(ylabel)
return fig
def matshow_tseries(time_series, fig=None, axis=0, xtick_n=5, time_unit=None,
xlabel=None, ylabel=None):
"""Creates an image of the time-series, ordered according to the first
dimension of the time-series object
Parameters
----------
time_series: a nitime time-series object
fig: a figure handle, opens a new figure if None
axis: an axis number (if there are several in the figure to be opened),
defaults to 0.
xtick_n: int, optional, sets the number of ticks to be placed on the x axis
"""
if fig is None:
fig = plt.figure()
if not fig.get_axes():
ax = fig.add_subplot(1, 1, 1)
else:
ax = fig.get_axes()[axis]
#Make sure that time displays on the x axis with the units you want:
#If you want to change the time-unit on the visualization from that used to
#represent the time-series:
if time_unit is not None:
tu = time_unit
conv_fac = ts.time_unit_conversion[time_unit]
#Otherwise, get the information from your input:
else:
tu = time_series.time_unit
conv_fac = time_series.time._conversion_factor
this_time = time_series.time / float(conv_fac)
ax.matshow(time_series.data)
ax.set_xticks(list(range(len(this_time)))[::len(this_time) / xtick_n])
ax.set_xticklabels(this_time[::len(this_time) / xtick_n])
if xlabel is None:
ax.set_xlabel('Time (%s)' % tu)
else:
ax.set_xlabel(xlabel)
if ylabel is not None:
ax.set_ylabel(ylabel)
return fig
def plsvfig(LLS,NNS,X,projval,x_label,y_label,z_label,Title_fnc,save_fnc):
fig = pl.figure()
ax = fig.gca(projection = '3d')
surf=ax.plot_surface(1e9*LS,NS,X,rstride=1,cstride=1,alpha=0.7,cmap=cm.winter,linewidth = 0.05, antialiased = True, shade = False)
CS = contour(1e9*LLS,NNS,X, colors = 'k', linewidth = 0.5)
cbar = pl.colorbar(surf)
#cbar.ax.set_ylabel(r'$\mathcal{A}^{(0)}_{[9,3]}\,\,\,[zJ]$', size = 14)
cbar.add_lines(CS)
cset = ax.contour(1e9*LLS,NNS,X,zdir='z',offset = projval ,cmap = cm.Blues)
ax.set_xlabel(x_label)
ax.set_ylabel(y_label)
ax.set_zlabel(z_label)
pl.title(Title_fnc)
#pl.title(r'$\rm{Gradient\, \mathcal{A}^{(0)}\, for \,[9,3]\, in\,water}$',size = 21)
ax.view_init(elev = 17, azim = 150)
savefig(save_fnc)#'plots/grad_A0_93_0.png')#, dpi = 300)
#pl.show()
return 0
def plot_xcorr(xc, ij, fig=None, line_labels=None, xticks=None, yticks=None,
xlabel=None, ylabel=None):
""" Visualize the cross-correlation function"""
if fig is None:
fig = plt.figure()
if not fig.get_axes():
ax = fig.add_subplot(1, 1, 1)
else:
ax = fig.get_axes()[0]
if line_labels is not None:
#Reverse the order, so that pop() works:
line_labels.reverse()
this_labels = line_labels
#Use the ij input as the labels:
else:
this_labels = [str(this) for this in ij].reverse()
#Make sure that time displays on the x axis with the units you want:
conv_fac = xc.time._conversion_factor
this_time = xc.time / float(conv_fac)
for (i, j) in ij:
if this_labels is not None:
#Use pop() to get the first one and remove it:
ax.plot(this_time, xc.data[i, j].squeeze(),
label=this_labels.pop())
else:
ax.plot(this_time, xc.data[i, j].squeeze())
ax.set_xlabel('Time(sec)')
ax.set_ylabel('Correlation(normalized)')
if xlabel is None:
#Make sure that time displays on the x axis with the units you want:
conv_fac = xc.time._conversion_factor
time_label = xc.time / float(conv_fac)
ax.set_xlabel('Time (%s)' % xc.time_unit)
else:
time_label = xlabel
ax.set_xlabel(xlabel)
if line_labels is not None:
plt.legend()
if ylabel is None:
ax.set_ylabel('Correlation')
else:
ax.set_ylabel(ylabel)
return fig
##pl.axis([0,1.0,0,1.0])
#pl.grid()
pl.savefig('plots/skew_ret_water/logG_contour.pdf')
show()
fig = pl.figure()
ax = fig.gca(projection = '3d')
#ax.text(-7, 6, 0.7, r'$\zeta/\omega_{0}$', zdir = (-1,1,-3), size = 21)
surf = ax.plot_surface(X,Y, G_l_t_dt, rstride = 1, cstride = 1,alpha = 0.2, linewidth = 0.3)#edgecolor = 'none',antialiased = True, shade = False, norm = norm, linewidth = 0.3)
#surf = ax.plot_surface(X,Y, G_l_t_dt, rstride = 20, cstride = 20,alpha = 0.2)#, cmap = cm.gnuplot, linewidth = 0.5)#gray)#coolwarm)#bone)#hot, linewidth = 0.01, antialiased = True, shade = False)# True)#, cmap = hot()
#colorbar(surf)
#cbar.ax.set_ylabel(r'$\frac{\xi}{\omega_{0}}$', size = 24)
#cset = ax.contour(X,Y,h, zdir = 'z', offset = 0, cmap = cm.jet)
#cset = ax.contour(X,Y,h, zdir = 'x', offset = 5, cmap = cm.jet)
#cset = ax.contourf(X,Y,h, zdir = 'y', offset = 6, cmap = cm.jet)# puts plot of max xi vs discrete r values at r=0 plane
#ax.view_init(elev = 19, azim = -112)
#zlabel(r'$\xi/\omega_{0}$', size = 21)
#ylabel(r'$r$', size = 24)
#xlabel(r'$(\epsilon(0) -1)$', size = 24)
#text = Axes.text(self, x, y, s, **kwargs)
#art3d.text_2d_to_3d(text, z, zdir)
#return text
#pl.text(6,0, 0, r'$\xi/\omega_{0}$',size = 21 ,rotation = 'horizontal')
#ax.text(r'$\xi/\omega_{0}$',6,0, 0, size = 21 ,rotation = 'horizontal')
#ax.set_zlabel(r'$\xi/\omega_{0}$',size = 21 ,rotation = 'horizontal' )
ax.set_xlabel(r'$\epsilon(0)-1$', size = 21)
ax.set_ylabel(r'$r$', size = 22)
show()
#pp.savefig()
def plot_tseries(time_series, fig=None, axis=0,
xticks=None, xunits=None, yticks=None, yunits=None,
xlabel=None, ylabel=None, yerror=None, error_alpha=0.1,
time_unit=None, **kwargs):
"""plot a timeseries object
Arguments
---------
time_series: a nitime time-series object
fig: a figure handle, opens a new figure if None
subplot: an axis number (if there are several in the figure to be opened),
defaults to 0.
xticks: optional, list, specificies what values to put xticks on. Defaults
to the matlplotlib-generated.
yticks: optional, list, specificies what values to put xticks on. Defaults
to the matlplotlib-generated.
xlabel: optional, list, specificies what labels to put on xticks
ylabel: optional, list, specificies what labels to put on yticks
yerror: optional, UniformTimeSeries with the same sampling_rate and number
of samples and channels as time_series, the error will be displayed as a
shading above and below the plotted time-series
"""
if fig is None:
fig = plt.figure()
if not fig.get_axes():
ax = fig.add_subplot(1, 1, 1)
else:
ax = fig.get_axes()[axis]
#Make sure that time displays on the x axis with the units you want:
#If you want to change the time-unit on the visualization from that used to
#represent the time-series:
if time_unit is not None:
tu = time_unit
conv_fac = ts.time_unit_conversion[time_unit]
#Otherwise, get the information from your input:
else:
tu = time_series.time_unit
conv_fac = time_series.time._conversion_factor
this_time = time_series.time / float(conv_fac)
ax.plot(this_time, time_series.data.T, **kwargs)
if xlabel is None:
ax.set_xlabel('Time (%s)' % tu)
else:
ax.set_xlabel(xlabel)
if ylabel is not None:
ax.set_ylabel(ylabel)
if yerror is not None:
if len(yerror.data.shape) == 1:
this_e = yerror.data[np.newaxis, :]
else:
this_e = yerror.data
delta = this_e
e_u = time_series.data + delta
e_d = time_series.data - delta
for i in range(e_u.shape[0]):
ax.fill_between(this_time, e_d[i], e_u[i], alpha=error_alpha)
return fig
def plot_corr_diff(tseries1, tseries2, fig=None,
ts_names=['1', '2']):
"""
Show the differences in *Fischer-transformed* snr correlations for two
time-series
Parameters
----------
tseries1, tseries2 : nitime TimeSeries objects
These are the time-series to compare, with each of them having the
dims: (n_channels, n_reps, time), where n_channels1 = n_channels2
lb,ub: float
Lower and upper bounds on the frequency range over which to
calculate the information rate (default to [0,Nyquist]).
fig: matplotlib figure object
If you want to do this on already existing figure. Otherwise, a new
figure object will be generated.
ts_names: list of str
Labels for the two inputs, to be used in plotting (defaults to
['1','2'])
bandwidth, adaptive, low_bias: See :func:`SNRAnalyzer` for details
Returns
-------
fig: a matplotlib figure object
"""
if fig is None:
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
SNR1 = []
SNR2 = []
corr1 = []
corr2 = []
corr_e1 = []
corr_e2 = []
for i in range(tseries1.shape[0]):
SNR1.append(nta.SNRAnalyzer(ts.TimeSeries(tseries1.data[i],
sampling_rate=tseries1.sampling_rate)))
corr1.append(np.arctanh(np.abs(SNR1[-1].correlation[0])))
corr_e1.append(SNR1[-1].correlation[1])
SNR2.append(nta.SNRAnalyzer(ts.TimeSeries(tseries2.data[i],
sampling_rate=tseries2.sampling_rate)))
corr2.append(np.arctanh(np.abs(SNR2[-1].correlation[0])))
corr_e2.append(SNR1[-1].correlation[1])
ax.scatter(np.array(corr1), np.array(corr2))
ax.errorbar(np.mean(corr1), np.mean(corr2),
yerr=np.std(corr2),
xerr=np.std(corr1))
plot_min = min(min(corr1), min(corr2))
plot_max = max(max(corr1), max(corr2))
ax.plot([plot_min, plot_max], [plot_min, plot_max], 'k--')
ax.set_xlabel('Correlation (Fischer Z) %s' % ts_names[0])
ax.set_ylabel('Correlation (Fischer Z) %s' % ts_names[1])
return fig, corr1, corr2
def plotTSs1(start_date, end_date, shortMa, longMa):
yf.pdr_override()
plt.close()
shortMaP = int(shortMa)
longMaP = int(longMa)
data = pdr.get_data_yahoo("BTC-USD", start_date, end_date)
btc = pd.DataFrame(data)
btc = btc[['Adj Close']]
btc.columns = ['price']
btc.reset_index(level=0, inplace=True)
###################################################################################
# Code taken from Willems, K., 2019. (Tutorial) Python For Finance: Algorithmic Trading
#[online] DataCamp Community. Available at: <https://www.datacamp.com/community/tutorials/finance-python-trading#basics>
plt.close()
signal = pd.DataFrame(index=btc.index) #copy index from BTC
signal['id'] = 0.0
signal['shortMA'] = btc['price'].rolling(
window=shortMaP,
min_periods=1).mean() #Short Moving Average Calculated
signal['longMA'] = btc['price'].rolling(
window=longMaP, min_periods=1).mean() ##Long Moving Average Calculated
signal['id'][shortMaP:] = np.where(
signal['shortMA'][shortMaP:] > signal['longMA'][shortMaP:], 1.0,
0.0) #where short moving average surpasses long, id is changed to 1
#[shortMaP:] only for the period that is surpasses than the short moving average
signal['buySell'] = signal['id'].diff()
fig = plt.figure()
ax1 = fig.add_subplot(111, ylabel=' Currency Price in $')
ax1.plot(btc.price) #Plot price
# Plot the short and long moving averages
ax1.plot(signal['shortMA'], color='black', label='Short Moving Average')
ax1.plot(signal['longMA'], color='orange', label='Long Moving Average')
ax1.legend(loc='upper right')
# The buy signals according to buySell are plotted
ax1.plot(signal.loc[signal.buySell == 1.0].index,
signal.shortMA[signal.buySell == 1.0],
'^',
markersize=10,
color='green')
# The sell signals according to buySell are plotted
ax1.plot(signal.loc[signal.buySell == -1.0].index,
signal.shortMA[signal.buySell == -1.0],
'v',
markersize=10,
color='red')
# Show the plot
plt.show()
###################################################################################
# Code taken from Willems, K., 2019. (Tutorial) Python For Finance: Algorithmic Trading
portfolio = pd.DataFrame(index=signal.index)
initInvestment = 100000
stocksOwned = pd.DataFrame(index=signal.index).fillna(0.0)
noCur = 10 #No of currency to be purchased
stocksOwned['BTC'] = noCur * signal['id']
portfolio['Holdings'] = stocksOwned['BTC'].multiply(btc['price'], axis=0)
buySell = stocksOwned['BTC'].diff()
portfolio['cash'] = initInvestment - (buySell.multiply(btc['price'],
axis=0)).cumsum()
portfolio['total'] = portfolio['cash'] + portfolio['Holdings']
portfolio['cash'][0] = initInvestment
portfolio['total'][0] = initInvestment
###################################################################################
fig = plt.figure()
ax = fig.add_subplot(111)
ax.plot(portfolio.index, portfolio['total'], label='Price')
ax.set_xlabel('Date')
ax.set_ylabel('Value of Portfolio in $')
day = portfolio.loc[signal.buySell == 1.0].index
day2 = portfolio.loc[signal.buySell == -1.0].index
#x co
ax.scatter(x=day, y=portfolio.loc[day, 'total'], color='green', marker='^')
ax.scatter(x=day2, y=portfolio.loc[day2, 'total'], color='red', marker='^')
surf=ax.plot_surface(1e9*LS,NS,X_91,rstride=1,cstride=1, alpha=0.5,cmap=cm.Greens,linewidth = 0.05, antialiased = True, shade = False)
surf=ax.plot_surface(1e9*LS,NS,X_93,rstride=1,cstride=1, alpha=0.5,cmap=cm.Reds,linewidth = 0.05, antialiased = True, shade = False)
surf=ax.plot_surface(1e9*LS,NS,X_290,rstride=1,cstride=1,alpha=0.5,cmap=cm.hot,linewidth = 0.05, antialiased = True, shade = False)
#cbar.ax.set_ylabel(r'$\mathcal{A}^{(0)}_{[9,3]}\,\,\,[zJ]$', size = 14)
#pl.title(r'$\rm{Gradient\, \mathcal{A}^{(0)}\, for \,[9,3]\, in\,water}$',size = 21)
ax.view_init(elev = 17, azim = 150)
#fig = pl.figure()
#ax = fig.gca(projection = '3d')
#surf=ax.plot_surface(1e9*LS,NS,X,rstride=1,cstride=1,alpha=0.7,cmap=cm.winter,linewidth = 0.05, antialiased = True, shade = False)
#CS = contour(1e9*LS,NS,X, colors = 'k', linewidth = 0.5)
#cbar = pl.colorbar(surf)
#cbar.ax.set_ylabel(r'$\mathcal{A}^{(0)}_{[9,3]}\,\,\,[zJ]$', size = 14)
#cbar.add_lines(CS)
#cset = ax.contour(1e9*LS,NS,X,zdir='z',offset = -0.25,cmap = cm.Blues)
ax.set_xlabel(r'$\rm{separation}\,\,\,\ell\,\,[nm]$', size = 18)
ax.set_ylabel(r'$\rm{n^{th}}\,Matsubra\,term$', size = 18)
ax.set_zlabel(r'$\nabla \mathcal{A}^{(0)}\,\,[zJ]$',size = 18 ,rotation = 'horizontal' )
#pl.title(r'$\rm{Gradient\, \mathcal{A}^{(0)}\, for \,[9,3]\, in\,water}$',size = 21)
#ax.view_init(elev = 17, azim = 10)
savefig('plots/grad_A0_combo.png')#, dpi = 300)
#show()
#
fig = pl.figure()
ax = fig.gca(projection = '3d')
contour(1e9*LS,NS,X_65, rstride=1,cstride=1,alpha=0.5,cmap=cm.Blues, linewidth = 0.05, antialiased = True, shade = False)
contour(1e9*LS,NS,X_91, rstride=1,cstride=1,alpha=0.5,cmap=cm.Greens, linewidth = 0.05, antialiased = True, shade = False)
contour(1e9*LS,NS,X_93, rstride=1,cstride=1,alpha=0.5,cmap=cm.Reds, linewidth = 0.05, antialiased = True, shade = False)
contour(1e9*LS,NS,X_290,rstride=1,cstride=1,alpha=0.5,cmap=cm.Purples,linewidth = 0.05, antialiased = True, shade = False)
#CS = contour(1e9*LS,NS,X, colors = 'k', linewidth = 0.5)
#cbar = pl.colorbar(surf)
#cbar.ax.set_ylabel(r'$\mathcal{A}^{(0)}_{[9,3]}\,\,\,[zJ]$', size = 14)
cbar_091_0.add_lines(CS_091_0)
cbar_290_0 = pl.colorbar(surf_290_0)
cbar_290_0.ax.set_ylabel(r'$\mathcal{A}^{(0)}_{[29,0]}\,\,\,[zJ]$', size = 14)
cbar_290_0.add_lines(CS_290_0)
cset_065_0 = ax.contour(1e9*X,Y,1e21*A0_065_theta,zdir='y',offset = -0.3,cmap = cm.Blues)
cset_091_0 = ax.contour(1e9*X,Y,1e21*A0_091_theta,zdir='y',offset = -0.3,cmap = cm.Greens)
cset_290_0 = ax.contour(1e9*X,Y,1e21*A0_290_theta,zdir='y',offset = -0.3,cmap = cm.Reds)
#man_loc = [(.1,.1),(.2,.2),(.3,.3),(.4,.4)]
yticks([0, pi/8, pi/6, pi/4, pi/3, pi/2],
['$0$', r'$\frac{\pi}{8}$', r'$\frac{\pi}{6}$', r'$\frac{\pi}{4}$', r'$\frac{\pi}{3}$', r'$\frac{\pi}{2}$'])
#clabel(CS, inline =1,fmt = '%1.1f', fontsize = 18,color = 'k', manual = man_loc)
ax.grid(on = True)
ax.set_xlabel(r'$\rm{separation}\,\,\,\ell\,\,[nm]$', size = 18)
ax.set_ylabel(r'$\rm{angle}\,\,\,\theta\,\,[radians]$', size = 18)
ax.set_zlabel(r'$\mathcal{A}^{(0)}\,\,[zJ]$',size = 18 ,rotation = 'horizontal' )
pl.title(r'$\rm{\mathcal{A}^{(0)}\, for \, [6,5],[9,1],\,and\,[29,0]\, in\,water}$',
size = 21)
ax.view_init(elev = 10, azim = 65)
savefig('plots/A0_65_91_290.png')#, dpi = 300)
show()
##### A_2 PLOTS ######
fig = figure()
ax = fig.gca(projection = '3d')
surf_065 = ax.plot_surface(1e9*X,Y,1e21*A2_065_theta, rstride = 5,
cstride =5,alpha=0.7,cmap=cm.Blues, linewidth = 0.05, antialiased = True, shade = False)# True)#, cmap = hot()
surf_091 = ax.plot_surface(1e9*X,Y,1e21*A2_091_theta, rstride = 5,
cstride =5,alpha=0.7,cmap=cm.Greens, linewidth = 0.05, antialiased = True, shade = False)# True)#, cmap = hot()
surf_290 = ax.plot_surface(1e9*X,Y,1e21*A2_290_theta, rstride = 5,
# pl.grid()
pl.savefig("plots/skew_NR_water/logG_contour.pdf")
show()
fig = pl.figure()
ax = fig.gca(projection="3d")
# ax.text(-7, 6, 0.7, r'$\zeta/\omega_{0}$', zdir = (-1,1,-3), size = 21)
surf = ax.plot_surface(
X, Y, G_l_t_dt, rstride=1, cstride=1, alpha=0.2, linewidth=0.3
) # edgecolor = 'none',antialiased = True, shade = False, norm = norm, linewidth = 0.3)
# surf = ax.plot_surface(X,Y, G_l_t_dt, rstride = 20, cstride = 20,alpha = 0.2)#, cmap = cm.gnuplot, linewidth = 0.5)#gray)#coolwarm)#bone)#hot, linewidth = 0.01, antialiased = True, shade = False)# True)#, cmap = hot()
# colorbar(surf)
# cbar.ax.set_ylabel(r'$\frac{\xi}{\omega_{0}}$', size = 24)
# cset = ax.contour(X,Y,h, zdir = 'z', offset = 0, cmap = cm.jet)
# cset = ax.contour(X,Y,h, zdir = 'x', offset = 5, cmap = cm.jet)
# cset = ax.contourf(X,Y,h, zdir = 'y', offset = 6, cmap = cm.jet)# puts plot of max xi vs discrete r values at r=0 plane
# ax.view_init(elev = 19, azim = -112)
# zlabel(r'$\xi/\omega_{0}$', size = 21)
# ylabel(r'$r$', size = 24)
# xlabel(r'$(\epsilon(0) -1)$', size = 24)
# text = Axes.text(self, x, y, s, **kwargs)
# art3d.text_2d_to_3d(text, z, zdir)
# return text
# pl.text(6,0, 0, r'$\xi/\omega_{0}$',size = 21 ,rotation = 'horizontal')
# ax.text(r'$\xi/\omega_{0}$',6,0, 0, size = 21 ,rotation = 'horizontal')
# ax.set_zlabel(r'$\xi/\omega_{0}$',size = 21 ,rotation = 'horizontal' )
ax.set_xlabel(r"$\epsilon(0)-1$", size=21)
ax.set_ylabel(r"$r$", size=22)
show()
# pp.savefig()
def plot_tseries(time_series,
fig=None,
axis=0,
xticks=None,
xunits=None,
yticks=None,
yunits=None,
xlabel=None,
ylabel=None,
yerror=None,
error_alpha=0.1,
time_unit=None,
**kwargs):
"""plot a timeseries object
Arguments
---------
time_series: a nitime time-series object
fig: a figure handle, opens a new figure if None
subplot: an axis number (if there are several in the figure to be opened),
defaults to 0.
xticks: optional, list, specificies what values to put xticks on. Defaults
to the matlplotlib-generated.
yticks: optional, list, specificies what values to put xticks on. Defaults
to the matlplotlib-generated.
xlabel: optional, list, specificies what labels to put on xticks
ylabel: optional, list, specificies what labels to put on yticks
yerror: optional, UniformTimeSeries with the same sampling_rate and number
of samples and channels as time_series, the error will be displayed as a
shading above and below the plotted time-series
"""
if fig is None:
fig = plt.figure()
if not fig.get_axes():
ax = fig.add_subplot(1, 1, 1)
else:
ax = fig.get_axes()[axis]
#Make sure that time displays on the x axis with the units you want:
#If you want to change the time-unit on the visualization from that used to
#represent the time-series:
if time_unit is not None:
tu = time_unit
conv_fac = ts.time_unit_conversion[time_unit]
#Otherwise, get the information from your input:
else:
tu = time_series.time_unit
conv_fac = time_series.time._conversion_factor
this_time = time_series.time / float(conv_fac)
ax.plot(this_time, time_series.data.T, **kwargs)
if xlabel is None:
ax.set_xlabel('Time (%s)' % tu)
else:
ax.set_xlabel(xlabel)
if ylabel is not None:
ax.set_ylabel(ylabel)
if yerror is not None:
if len(yerror.data.shape) == 1:
this_e = yerror.data[np.newaxis, :]
else:
this_e = yerror.data
delta = this_e
e_u = time_series.data + delta
e_d = time_series.data - delta
for i in range(e_u.shape[0]):
ax.fill_between(this_time, e_d[i], e_u[i], alpha=error_alpha)
return fig
def plot_xcorr(xc,
ij,
fig=None,
line_labels=None,
xticks=None,
yticks=None,
xlabel=None,
ylabel=None):
""" Visualize the cross-correlation function"""
if fig is None:
fig = plt.figure()
if not fig.get_axes():
ax = fig.add_subplot(1, 1, 1)
else:
ax = fig.get_axes()[0]
if line_labels is not None:
#Reverse the order, so that pop() works:
line_labels.reverse()
this_labels = line_labels
#Use the ij input as the labels:
else:
this_labels = [str(this) for this in ij].reverse()
#Make sure that time displays on the x axis with the units you want:
conv_fac = xc.time._conversion_factor
this_time = xc.time / float(conv_fac)
for (i, j) in ij:
if this_labels is not None:
#Use pop() to get the first one and remove it:
ax.plot(this_time,
xc.data[i, j].squeeze(),
label=this_labels.pop())
else:
ax.plot(this_time, xc.data[i, j].squeeze())
ax.set_xlabel('Time(sec)')
ax.set_ylabel('Correlation(normalized)')
if xlabel is None:
#Make sure that time displays on the x axis with the units you want:
conv_fac = xc.time._conversion_factor
time_label = xc.time / float(conv_fac)
ax.set_xlabel('Time (%s)' % xc.time_unit)
else:
time_label = xlabel
ax.set_xlabel(xlabel)
if line_labels is not None:
plt.legend()
if ylabel is None:
ax.set_ylabel('Correlation')
else:
ax.set_ylabel(ylabel)
return fig
def plot_corr_diff(tseries1, tseries2, fig=None, ts_names=['1', '2']):
"""
Show the differences in *Fischer-transformed* snr correlations for two
time-series
Parameters
----------
tseries1, tseries2 : nitime TimeSeries objects
These are the time-series to compare, with each of them having the
dims: (n_channels, n_reps, time), where n_channels1 = n_channels2
lb,ub: float
Lower and upper bounds on the frequency range over which to
calculate the information rate (default to [0,Nyquist]).
fig: matplotlib figure object
If you want to do this on already existing figure. Otherwise, a new
figure object will be generated.
ts_names: list of str
Labels for the two inputs, to be used in plotting (defaults to
['1','2'])
bandwidth, adaptive, low_bias: See :func:`SNRAnalyzer` for details
Returns
-------
fig: a matplotlib figure object
"""
if fig is None:
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1)
SNR1 = []
SNR2 = []
corr1 = []
corr2 = []
corr_e1 = []
corr_e2 = []
for i in range(tseries1.shape[0]):
SNR1.append(
nta.SNRAnalyzer(
ts.TimeSeries(tseries1.data[i],
sampling_rate=tseries1.sampling_rate)))
corr1.append(np.arctanh(np.abs(SNR1[-1].correlation[0])))
corr_e1.append(SNR1[-1].correlation[1])
SNR2.append(
nta.SNRAnalyzer(
ts.TimeSeries(tseries2.data[i],
sampling_rate=tseries2.sampling_rate)))
corr2.append(np.arctanh(np.abs(SNR2[-1].correlation[0])))
corr_e2.append(SNR1[-1].correlation[1])
ax.scatter(np.array(corr1), np.array(corr2))
ax.errorbar(np.mean(corr1),
np.mean(corr2),
yerr=np.std(corr2),
xerr=np.std(corr1))
plot_min = min(min(corr1), min(corr2))
plot_max = max(max(corr1), max(corr2))
ax.plot([plot_min, plot_max], [plot_min, plot_max], 'k--')
ax.set_xlabel('Correlation (Fischer Z) %s' % ts_names[0])
ax.set_ylabel('Correlation (Fischer Z) %s' % ts_names[1])
return fig, corr1, corr2
def plotTsS2(start_date, end_date):
plt.close()
yf.pdr_override()
data = pdr.get_data_yahoo("BTC-USD", start_date, end_date)
btc = pd.DataFrame(data)
btc = btc[['Adj Close']]
btc.columns = ['Close']
btc.reset_index(level=0, inplace=True)
b = pd.DataFrame()
for x in btc:
b['price'] = btc['Close']
b['sma'] = btc['Close'].rolling(window=20).mean()
b['std'] = btc['Close'].rolling(window=20).std()
b['bolU'] = b['sma'] + (2 * b['std']) #Calculating Upper Bound
b['bolD'] = b['sma'] - (2 * b['std']) #Calculating Lower Bound
#Convert Bollinger Bands to %b - bollinger column
b['bollinger'] = (b['price'] - b['bolD']) / (b['bolU'] - b['bolD'])
bb1 = b[['price', 'bolU', 'bolD', 'bollinger']]
bb1.columns = ['Price', 'Upper Band', 'Lower Band', 'Bollinger']
bb1.fillna(0, inplace=True)
# In[ ]:
# In[9]:
RSI = pd.DataFrame(index=btc.index)
RSI['price'] = btc['Close']
RSI['val'] = None
RSI['up'] = 0 #When there is no up value 0
RSI['down'] = 0 #When there is no down value 0
size = RSI.shape[0]
dp = 14
for x in range(size):
if x == 0:
continue #first day will continue
#calculating the ups , when closing price is higher in day x than x -1
if RSI['price'].iloc[x] > RSI['price'].iloc[x - 1]: #
RSI['up'].iloc[x] = RSI['price'].iloc[x] - RSI['price'].iloc[x - 1]
else:
#calculating the downs days , when closing price is lower in day x than x -1
RSI['down'].iloc[x] = RSI['price'].iloc[x -
1] - RSI['price'].iloc[x]
if x >= dp:
avgUp = RSI['up'][x - dp:x].sum(
) / dp #calculates avg up of last dp days
avgDown = RSI['down'][x - dp:x].sum(
) / dp #calculates avg down of last dp days
rs = avgUp / avgDown #calculation of RS
RSI['val'].iloc[x] = 100 - 100 / (1 + rs)
signals = pd.DataFrame(index=btc.index) #copy index for BTC
signals['price'] = btc['Close']
signals['id'] = 0.0
signals['RSI'] = RSI['val']
signals['RSI'].fillna(0, inplace=True)
signals['bollinger'] = bb1['Bollinger']
signals['id'] = [np.nan for i in signals.index]
# only verifications for days after DPth (period of RSI) day
signals['id'][dp:].loc[((signals['RSI'] < 30) &
(signals['bollinger'] < 0))] = 1
signals['id'][dp:].loc[((signals['RSI'] > 70) &
(signals['bollinger'] > 1))] = 0
signals['id'].ffill(inplace=True) #fill empty values with 0
signals['id'].fillna(0, inplace=True)
signals['buySell'] = signals['id'].diff()
signals['buySell'].fillna(0, inplace=True)
fig, (ax) = plt.subplots(1, 1)
ax.plot(signals['price'])
ax.set_xlabel('Date')
ax.set_ylabel('Value in $')
day = signals.loc[signals.buySell == 1.0].index
day2 = signals.loc[signals.buySell == -1.0].index
#x cordinates, index of day --> y cordinate,price of day index
ax.scatter(
x=day,
y=signals.loc[day, 'price'],
marker='^',
color='green',
)
ax.scatter(x=day2, y=signals.loc[day2, 'price'], marker='v', color='red')
plt.show()
def plotPorts2(start_date, end_date):
plt.close()
yf.pdr_override()
data = pdr.get_data_yahoo("BTC-USD", start_date, end_date)
btc = pd.DataFrame(data)
btc = btc[['Adj Close']]
btc.columns = ['Close']
btc.reset_index(level=0, inplace=True)
# In[8]:
b = pd.DataFrame()
for x in btc:
b['price'] = btc['Close']
b['sma'] = btc['Close'].rolling(window=20).mean()
b['std'] = btc['Close'].rolling(window=20).std()
b['bolU'] = b['sma'] + (2 * b['std']) #Calculating Upper Bound
b['bolD'] = b['sma'] - (2 * b['std']) #Calculating Lower Bound
#Convert Bollinger Bands to %b - bollinger column
b['bollinger'] = (b['price'] - b['bolD']) / (b['bolU'] - b['bolD'])
bb1 = b[['price', 'bolU', 'bolD', 'bollinger']]
bb1.columns = ['Price', 'Upper Band', 'Lower Band', 'Bollinger']
bb1.fillna(0, inplace=True)
# In[ ]:
# In[9]:
RSI = pd.DataFrame(index=btc.index)
RSI['price'] = btc['Close']
RSI['val'] = None
RSI['up'] = 0 #When there is no up value 0
RSI['down'] = 0 #When there is no down value 0
size = RSI.shape[0]
dp = 14
for x in range(size):
if x == 0:
continue #first day will continue
#calculating the ups , when closing price is higher in day x than x -1
if RSI['price'].iloc[x] > RSI['price'].iloc[x - 1]: #
RSI['up'].iloc[x] = RSI['price'].iloc[x] - RSI['price'].iloc[x - 1]
else:
#calculating the downs days , when closing price is lower in day x than x -1
RSI['down'].iloc[x] = RSI['price'].iloc[x -
1] - RSI['price'].iloc[x]
if x >= dp:
avgUp = RSI['up'][x - dp:x].sum(
) / dp #calculates avg up of last dp days
avgDown = RSI['down'][x - dp:x].sum(
) / dp #calculates avg down of last dp days
rs = avgUp / avgDown #calculation of RS
RSI['val'].iloc[x] = 100 - 100 / (1 + rs)
signals = pd.DataFrame(index=btc.index) #copy index for BTC
signals['price'] = btc['Close']
signals['id'] = 0.0
signals['RSI'] = RSI['val']
signals['RSI'].fillna(0, inplace=True)
signals['bollinger'] = bb1['Bollinger']
signals['id'] = [np.nan for i in signals.index]
# only verifications for days after DPth (period of RSI) day
signals['id'][dp:].loc[((signals['RSI'] < 30) &
(signals['bollinger'] < 0))] = 1
signals['id'][dp:].loc[((signals['RSI'] > 70) &
(signals['bollinger'] > 1))] = 0
signals['id'].ffill(inplace=True) #fill empty values with 0
signals['id'].fillna(0, inplace=True)
signals['buySell'] = signals['id'].diff()
signals['buySell'].fillna(0, inplace=True)
###################################################################################
# Code taken from Willems, K., 2019. (Tutorial) Python For Finance: Algorithmic Trading
initInvestment = 100000
stocksOwned = pd.DataFrame(index=signals.index).fillna(0.0)
noCur = 10 #No of currency to be purchased
stocksOwned['BTC'] = noCur * signals['id']
portfolio = pd.DataFrame(index=signals.index)
portfolio['Holdings'] = stocksOwned['BTC'].multiply(btc['Close'], axis=0)
buySell = stocksOwned['BTC'].diff()
portfolio['cash'] = initInvestment - (buySell.multiply(btc['Close'],
axis=0)).cumsum()
portfolio['total'] = portfolio['cash'] + portfolio['Holdings']
portfolio['cash'][0] = initInvestment
portfolio['total'][0] = initInvestment
###################################################################################
# In[50]:
fig, (ax) = plt.subplots(1, 1, sharex=True)
ax.plot(portfolio.index, portfolio['total'], label='Price')
ax.set_xlabel('Date')
ax.set_ylabel('Value of portfolio in USD')
day = signals.loc[signals.buySell == 1.0].index
day2 = signals.loc[signals.buySell == -1.0].index
ax.scatter(x=day, y=portfolio.loc[day, 'total'], color='green')
ax.scatter(x=day2, y=portfolio.loc[day2, 'total'], color='red')
plt.show()