def plot_Clusters_HMM_time(Xdata, Ks_params):
# this function takes the list of trials and computes the likelihoods
# of every sample of every trial belonging to a cluster, sums them up and
# then plots them
Ntrials = len(Xdata)
Nclasses = len(Ks_params)
total_llresp = []
for K_p_i in range(Nclasses):
total_llresp.append([0])
for trial_i in range(Ntrials):
for K_p_i in range(Nclasses): # For every cluster type
# print Xdata[trial_i].shape
A = Ks_params[K_p_i][1]
B = Ks_params[K_p_i][2]
pi = Ks_params[K_p_i][0]
alpha = HMMlf.get_alfa_matrix_log(A, B, pi, [Xdata[trial_i]])
beta = HMMlf.get_beta_matrix_log(A, B, pi, [Xdata[trial_i]])
ll_resp = HMMlf.get_gamma_matrix_log(alpha, beta)[0]
# print ll_resp[0].shape
ll_resp = np.exp(ll_resp).T
# for nsample
total_llresp[K_p_i] += ll_resp
# total_llresp[K_p_i] = HMMlf.sum_logs([total_llresp[K_p_i],ll_resp])
for K_p_i in range(Nclasses):
total_llresp[K_p_i] = total_llresp[K_p_i] / Ntrials
# total_llresp[0] = np.log (total_llresp[0])
# total_llresp[1] = np.log (total_llresp[1])
# Now we have the sum of log responsabilities !!
# Product for chains !!
# gl.plot([],total_llresp[0])
# gl.plot([],total_llresp[1])
for K_p_i in range(Nclasses):
pi = Ks_params[K_p_i][0]
NK = pi.size
labels = []
for k in range(NK):
labels.append("Cluster " + str(k + 1))
gl.plot_filled(
[],
total_llresp[K_p_i],
labels=["Responsability HMM", "Time index", "Responsability"],
legend=labels)
def plot_Clusters_HMM_time(Xdata,Ks_params):
# this function takes the list of trials and computes the likelihoods
# of every sample of every trial belonging to a cluster, sums them up and
# then plots them
Ntrials = len(Xdata)
Nclasses = len (Ks_params)
total_llresp = []
for K_p_i in range(Nclasses):
total_llresp.append([0])
for trial_i in range(Ntrials):
for K_p_i in range(Nclasses): # For every cluster type
# print Xdata[trial_i].shape
A = Ks_params[K_p_i][1]
B = Ks_params[K_p_i][2]
pi = Ks_params[K_p_i][0]
alpha = HMMlf.get_alfa_matrix_log(A,B,pi, [Xdata[trial_i]])
beta = HMMlf.get_beta_matrix_log(A,B,pi, [Xdata[trial_i]])
ll_resp = HMMlf.get_gamma_matrix_log(alpha, beta)[0]
# print ll_resp[0].shape
ll_resp = np.exp(ll_resp).T
# for nsample
total_llresp[K_p_i] += ll_resp
# total_llresp[K_p_i] = HMMlf.sum_logs([total_llresp[K_p_i],ll_resp])
for K_p_i in range(Nclasses):
total_llresp[K_p_i] = total_llresp[K_p_i]/Ntrials
# total_llresp[0] = np.log (total_llresp[0])
# total_llresp[1] = np.log (total_llresp[1])
# Now we have the sum of log responsabilities !!
# Product for chains !!
# gl.plot([],total_llresp[0])
# gl.plot([],total_llresp[1])
for K_p_i in range(Nclasses):
pi = Ks_params[K_p_i][0]
NK = pi.size
labels = []
for k in range(NK):
labels.append("Cluster " + str(k +1))
gl.plot_filled([],total_llresp[K_p_i], labels = ["Responsability HMM", "Time index","Responsability"], legend = labels)
"""
gl.init_figure()
nf = 0
AxesStyle = "Normal - No xaxis - Ny:3"
for i in range(days_plot):
chain = Xdata[i]
r = myEM.get_responsibilities([chain],myDManager, theta_list[-1],model_theta_list[-1])
if (myEM.clusters_relation == "independent"):
resp = r
elif(myEM.clusters_relation == "MarkovChain1"):
resp = r[0]
if (i == days_plot -1):
AxesStyle = "Normal - Ny:3"
ax1 = gl.plot_filled(Xdata_dates[i],resp , nf = nf, fill_mode = "stacked", legend = legend,
sharex = ax1, sharey = ax1, AxesStyle = AxesStyle, labels = [labels_title,"","D(%i)"%i],
dataTransform = dataTransform, step_mode = "yes")
gl.set_fontSizes(ax = ax1, title = 20, xlabel = 20, ylabel = 10,
legend = 20, xticks = 10, yticks = 10)
# ax1 = gl.step(Xdata_dates[i],resp , nf = nf, fill= 1, legend = legend,
# sharex = ax1, sharey = ax1, AxesStyle = "Normal", labels = [labels_title,"","%i"%i],
# dataTransform = dataTransform, alpha = 0.5)
gl.colorIndex = 0
labels_title = ""
gl.subplots_adjust(left=.09, bottom=.20, right=.90, top=.95, wspace=.2, hspace=0.001)
image_name = "EM_1Symbol_timeAnalysis"+ clusters_relation+ "_"+str(periods[0])+ '.png'
gl.savefig(folder_images + image_name,
##### but fed directly, we need to still make such a transformation for this to work?
gl.barchart(dates,
dataHLOC,
ax=ax1,
color=color1,
labels=[title, "", r"Price ($\$$)"],
legend=["HLOC Price"])
BBs = np.concatenate([BB1[:, [0]], BB1[:, [1]]], axis=1)
gl.plot(dates, SMA1, color=colorBB1, legend=["0.95 BB(%i)" % nBB1])
gl.plot(dates, BBs, color=colorBB1, ls="--")
gl.plot_filled(
dates,
BBs,
color=colorBB1,
alpha=0.1,
AxesStyle="Normal - No xaxis",
)
gl.plot(dates, BBs, color=colorBB1, ls="--")
gl.plot_filled(
dates,
BBs,
color=colorBB1,
alpha=0.1,
AxesStyle="Normal - No xaxis",
)
##
ax2 = gl.subplot2grid((5, 1), (4, 0), rowspan=1, colspan=1, sharex=ax1)
SMA2_ATR = timeData.EMA(n=nCha2)
## Plotting just the ATR !!
# TODO: Divide 3-1
gl.set_subplots(3, 1)
gl.plot(dates,
price,
nf=1,
labels=["Volatility indicator Chainkin", "", "Price"],
legend=["Price"])
gl.plot(dates, SMA1_ATR, nf=0, color="k", legend=["Chainkin(%i)" % nATR1])
gl.plot_filled(dates,
np.concatenate(
[SMA1_ATR + 2 * EMA_Range1, SMA1_ATR - 2 * EMA_Range1],
axis=1),
alpha=0.5,
nf=0,
color="k")
gl.plot(dates, SMA2_ATR, nf=0, color="y", legend=["Chainkin(%i)" % nATR2])
gl.plot_filled(dates,
np.concatenate(
[SMA2_ATR + 2 * EMA_Range2, SMA2_ATR - 2 * EMA_Range2],
axis=1),
alpha=0.5,
nf=0,
color="y")
gl.plot(dates,
# r = myEM.get_alpha_responsibilities([chain],myDManager, theta_list[-1],model_theta_list[-1])
if (myEM.clusters_relation == "independent"):
resp = r
elif (myEM.clusters_relation == "MarkovChain1"):
resp = r[0]
Nclusters = resp.shape[1]
ax_ii = gl.subplot2grid((days_plot, Ndiv), (i, 0),
rowspan=1,
colspan=Ndiv - 1)
ax1 = gl.plot_filled(Xdata_dates[i],
resp,
nf=0,
fill_mode="stacked",
legend=legend,
sharex=ax1,
sharey=ax1,
AxesStyle="Normal - No xaxis - No yaxis",
labels=[labels_title, "",
"%i" % i],
dataTransform=dataTransform,
step_mode="yes")
gl.colorIndex = 0
labels_title = ""
axes_l.append(ax_ii)
gl.subplots_adjust(left=.09,
bottom=.20,
right=.90,
top=.95,
wspace=.2,
## Plotting t the BB !!
ax1 = gl.subplot2grid((5,1), (0,0), rowspan=4, colspan=1)
# gl.plot(dates, price, ax = ax1, AxesStyle = "Normal - No xaxis",
# labels = ["Volatility indicator BB", "", "Price"],
# legend = ["Price"], color = "dark navy blue")
title = "Bollinger Bands. " + str(symbols[0]) + "(" + ul.period_dic[timeData.period]+ ")"
gl.barchart(dates, dataHLOC, ax = ax1, color = color1,
labels = [title, "", r"Price ($\$$)"],
legend = ["HLOC Price"])
BBs = np.concatenate([BB1[:,[0]],BB1[:,[1]]],axis = 1)
gl.plot(dates, SMA1, color = colorBB1,
legend = ["0.95 BB(%i)"%nBB1])
gl.plot(dates, BBs,
color = colorBB1, ls = "--")
gl.plot_filled(dates, BBs,
color = colorBB1, alpha = 0.1, AxesStyle = "Normal - No xaxis",)
gl.plot(dates, BBs,
color = colorBB1, ls = "--")
gl.plot_filled(dates, BBs,
color = colorBB1, alpha = 0.1, AxesStyle = "Normal - No xaxis",)
##
ax2 = gl.subplot2grid((5,1), (4,0), rowspan=1, colspan=1, sharex = ax1)
gl.plot(dates, BB1[:,0] - BB1[:,1] ,ax = ax2, color = colorBB1,
labels = ["","","STD"], AxesStyle = "Normal - Ny:5",
legend = ["BBstd(%i)"%nBB1], fill = 1, alpha = 0.5)
gl.subplots_adjust(left=.09, bottom=.10, right=.90, top=.95, wspace=.20, hspace=0)
for i in range(Npeople_plot):
if (myEM.clusters_relation == "independent"):
resp = r[N * i:N * (i + 1), :]
elif (myEM.clusters_relation == "MarkovChain1"):
resp = r[i]
Nclusters = resp.shape[1]
ax_ii = gl.subplot2grid((Npeople_plot, Ndiv), (i, 0),
rowspan=1,
colspan=Ndiv - 1)
ax1 = gl.plot_filled(time,
resp,
nf=0,
fill_mode="stacked",
legend=legend,
sharex=ax1,
sharey=ax1,
AxesStyle="Normal - No yaxis",
labels=[labels_title, "",
"%i" % i])
gl.colorIndex = 0
labels_title = ""
ax_i = gl.subplot2grid((Npeople_plot, Ndiv), (0, Ndiv - 1),
rowspan=Npeople_plot / 2,
colspan=1)
for i in range(1, K + 1):
gl.scatter(0,
i,
legend=[" K = %i" % (i)],
lw=28,
Ndiv = 4
axes_l = []
for i in range(days_plot):
chain = Xdata[i]
r = myEM.get_responsibilities([chain],myDManager, theta_list[-1],model_theta_list[-1])
# r = myEM.get_alpha_responsibilities([chain],myDManager, theta_list[-1],model_theta_list[-1])
if (myEM.clusters_relation == "independent"):
resp = r
elif(myEM.clusters_relation == "MarkovChain1"):
resp = r[0]
Nclusters = resp.shape[1]
ax_ii = gl.subplot2grid((days_plot,Ndiv), (i,0), rowspan=1, colspan=Ndiv-1)
ax1 = gl.plot_filled(Xdata_dates[i],resp , nf = 0, fill_mode = "stacked", legend = legend,
sharex = ax1, sharey = ax1, AxesStyle = "Normal - No xaxis - No yaxis", labels = [labels_title,"","%i"%i],
dataTransform = dataTransform, step_mode = "yes")
gl.colorIndex = 0
labels_title = ""
axes_l.append(ax_ii)
gl.subplots_adjust(left=.09, bottom=.20, right=.90, top=.95, wspace=.2, hspace=0.001)
ax_i = gl.subplot2grid((days_plot,Ndiv), (0,Ndiv-1), rowspan=int(days_plot/2), colspan=1)
for i in range(1,K+1):
gl.scatter(0, i, legend = [" K = %i"%(i)], lw = 28, AxesStyle = "Normal - No xaxis - No yaxis" , loc = "center left")
gl.subplots_adjust(left=.09, bottom=.10, right=.90, top=.95, wspace=.2, hspace=0.01)
image_name = "EM_%iSymbol_timeAnalysis_%iclusters"%(2,K)+ clusters_relation+ "_"+str(periods[0])+ '.png'
#
gl.set_fontSizes(ax = axes_l, title = 20, xlabel = 20, ylabel = 20,
for i in range(days_plot):
chain = Xdata[i]
r = myEM.get_responsibilities([chain], myDManager, theta_list[-1],
model_theta_list[-1])
if (myEM.clusters_relation == "independent"):
resp = r
elif (myEM.clusters_relation == "MarkovChain1"):
resp = r[0]
if (i == days_plot - 1):
AxesStyle = "Normal - Ny:3"
ax1 = gl.plot_filled(Xdata_dates[i],
resp,
nf=nf,
fill_mode="stacked",
legend=legend,
sharex=ax1,
sharey=ax1,
AxesStyle=AxesStyle,
labels=[labels_title, "",
"D(%i)" % i],
dataTransform=dataTransform,
step_mode="yes")
gl.set_fontSizes(ax=ax1,
title=20,
xlabel=20,
ylabel=10,
legend=20,
xticks=10,
yticks=10)
# ax1 = gl.step(Xdata_dates[i],resp , nf = nf, fill= 1, legend = legend,
# sharex = ax1, sharey = ax1, AxesStyle = "Normal", labels = [labels_title,"","%i"%i],