def computeAndPlotCDFforVibUse(analysis, n, end):
#computes cdf for number of days people used vibration on the device
earlyDate='2015-08-31' #assume data from users is good after this date
#assume users are not lost until cutOff number of days after last use
cutOff=10
userCount_old=numpy.zeros((n,1))
new_userCount=numpy.zeros((cutOff,1))
for ch in analysis.values():
if ch.start_date!=None and ch.end_date>earlyDate:
#print ch.end_date, end
if dateDiff(ch.start_date, end)>=cutOff:
#if ch.start_date>'2015-09-01':
index=len(ch.vibration_use)
userCount_old[index]+=1
else:
index=len(ch.vibration_use)
if index>cutOff:
print "problem with vibration use logic", ch.start_date
new_userCount[index]+=1
elif len(ch.vibration_use)>0:
print 'issue with data, id=', ch.child_id
else:
index=len(ch.vibration_use)
userCount_old[index]+=1
print 'number of old users using vibration', sum(userCount_old)
cdf_old=numpy.zeros((n,1))
cdf_new=numpy.zeros((cutOff,1))
for i in range(1,n):
cdf_old[i]=sum(userCount_old[:i])
for i in range(1, cutOff):
cdf_new[i]=sum(new_userCount[:i])
diff_old=numpy.zeros((n,1))
for i in range(1,n):
diff_old[i]=cdf_old[i]-cdf_old[i-1]
diff_new=numpy.zeros((cutOff,1))
for i in range(1, cutOff):
diff_new[i]=cdf_new[i]-cdf_new[i-1]
print 'new users', userCount_old[-1], 'old diff', numpy.transpose(diff_old[:50])
print 'new diff', numpy.transpose(diff_new)
plt.figure()
plt.bar(range(n), cdf_old/sum(userCount_old), align='center')
plt.title('cdf: number of people who stopped using the device')
plt.xlabel('days')
plt.ylabel('% of all users')
plt.figure()
plt.plot(diff_old, 'g*--')
plt.title('number of people lost for >'+str(cutOff))
plt.ylabel('number of people')
plt.xlabel('days')
plt.show()
def plotDevRateVsImprov(analysis):
#plots dev rate vs biggest improvement in dist event rate
improvement=[]
reactgood_rate=[]
for ch in analysis.values():
if len(ch.biggest_improvement)>0:
improvement.append(ch.biggest_improvement[0])
reactgood_rate.append(ch.devRate['ReactGood'])
plt.figure()
plt.plot(reactgood_rate, improvement, 'gs')
plt.title('sussessful device use rate vs improvement')
plt.ylabel('improvement in distuption events')
plt.xlabel('rate of correct reaction')
plt.show()
def plotUseRateVsBestRate(analysis):
#shows plot of use rate vs best dist event rate from biggest improvement
use_rate=[]
best_rate=[]
for ch in analysis.values():
if len(ch.use)>20:
if len(ch.biggest_improvement):
use_rate.append(ch.use_rate)
best_rate.append(ch.biggest_improvement[2])
plt.figure()
plt.title('best rate vs use freq in 4 weeks prior to best rate')
plt.xlabel('best rate')
plt.ylabel('use freq in 4 weeks')
plt.plot(best_rate, use_rate, 'bo')
plt.show()
def plotCorrVsOffset(analysis):
#plots correlation between dist event to reactgood and offset for start of company's device
correlation=[]
std_offset=[]
mean_offset=[]
for ch in analysis.values():
if len(ch.use)>0:
reactgood= ch.corr_dev_reactgood
std_off=ch.std_offset_for_device_start
mean=ch.avg_offset_for_device_start
if reactgood!=None and not numpy.isnan(std_off):
correlation.append(reactgood)
std_offset.append(std_off)
mean_offset.append(mean)
plt.figure()
plt.plot(correlation, std_offset, 'b*')
plt.figure()
plt.plot(correlation, mean_offset, 'r*')
plt.figure()
plt.plot(mean_offset, std_offset, 'g*')
plt.show()
def plotUseRateVsImprovRate(analysis):
#shows plot of use rate vs biggest improvement diff rate for distuption events
use_rate=[]
improv_diff=[]
best_rate=[]
for ch in analysis.values():
if len(ch.use)>20:
if len(ch.biggest_improvement):
use_rate.append(ch.use_rate)
improv_diff.append(ch.biggest_improvement[0])
best_rate.append(ch.biggest_improvement[2])
imdiffLp6=[]
imdiffBp6=[]
brateLp6=[]
brateBp6=[]
for i in range(len( use_rate)):
if use_rate[i]<.6:
imdiffLp6.append(improv_diff[i])
brateLp6.append(best_rate[i])
else:
imdiffBp6.append(improv_diff[i])
brateBp6.append(best_rate[i])
print '******Report 2******'
print 'for users with use rate <.6, improvement stats: mean', numpy.mean(imdiffLp6), 'std', numpy.std(imdiffLp6)
print 'for users with use rate <.6, best rate stats: mean', numpy.mean(brateLp6), 'std', numpy.std(brateLp6)
print 'for users with use rate >=.6, improvement stats: mean', numpy.mean(imdiffBp6), 'std', numpy.std(imdiffBp6)
print 'for users with use rate >=.6, best rate stats: mean', numpy.mean(brateBp6), 'std', numpy.std(brateBp6)
plt.figure()
plt.title('improvement difference vs use freq in 4 weeks prior to best rate')
plt.xlabel('improvement difference')
plt.ylabel('use freq in 4 weeks')
plt.plot(improv_diff, use_rate, 'bo')
plt.show()
def computeAndPlotCDF(analysis, n, end):
#computes cdf for number of days people used the app
userCount=numpy.zeros((n,1))
for ch in analysis.values():
index=len(ch.use)
userCount[index]+=1
cdf=numpy.zeros((n,1))
for i in range(1,n):
cdf[i]=sum(userCount[:i])
diff=numpy.zeros((n,1))
for i in range(1,n):
diff[i]=cdf[i]-cdf[i-1]
print 'all users', numpy.transpose(diff[:50])
print 'max', max(cdf), sum(userCount)
plt.figure()
plt.bar(range(n), cdf*1.0/sum(userCount), align='center')
plt.title('cdf: number of people who stopped using the app')
plt.xlabel('days')
plt.ylabel('% of all users')
plt.figure()
plt.plot(diff, 'g*--')
plt.title('number of people lost')
plt.ylabel('number of people')
plt.xlabel('days')
#assume users are not lost until "cutOff" days after last use
cutOff=10
userCount_old=numpy.zeros((n,1))
new_userCount=numpy.zeros((cutOff,1))
for ch in analysis.values():
if ch.start_date!=None:
#print ch.end_date, end
if dateDiff(ch.start_date, end)>=cutOff:
if ch.start_date>'2015-08-31':
index=len(ch.use)
userCount_old[index]+=1
else:
userCount_old[-1]+=1
index=len(ch.use)
if index>10:
print ch.start_date
new_userCount[index]+=1
elif len(ch.use)>0:
print 'issue with data, id=', ch.child_id
else:
index=len(ch.use)
userCount_old[index]+=1
print 'number of all users', len(analysis), 'number of old users', sum(userCount_old), sum(userCount)
cdf_old=numpy.zeros((n,1))
cdf_new=numpy.zeros((cutOff,1))
for i in range(1,n):
cdf_old[i]=sum(userCount_old[:i])
for i in range(1, cutOff):
cdf_new[i]=sum(new_userCount[:i])
diff_old=numpy.zeros((n,1))
for i in range(1,n):
diff_old[i]=cdf_old[i]-cdf_old[i-1]
diff_new=numpy.zeros((cutOff,1))
for i in range(1, cutOff):
diff_new[i]=cdf_new[i]-cdf_new[i-1]
print 'new users', userCount_old[-1], 'old diff', numpy.transpose(diff_old[:50])
print 'new diff', numpy.transpose(diff_new)
plt.figure()
plt.bar(range(n), cdf_old/sum(userCount_old), align='center')
plt.title('cdf: number of people who stopped using device for >'+str(cutOff))
plt.xlabel('days')
plt.ylabel('number of people')
plt.figure()
#plt.plot(range(n), 1-cdf*1.0/sum(userCount_old),'b-', ls='steps',linewidth=4.0)
plt.plot(range(n), 1-cdf_old/sum(userCount_old),'b-',ls='steps', linewidth=4.0)
plt.title('Remaining Users')
plt.xlabel('days')
plt.xlim((0,50))
plt.ylabel('fraction of all users')
#matplotlib.rcParams.update({'font.size': 22})
plt.figure()
plt.plot(diff_old, 'g*--')
plt.title('number of people lost for >'+str(cutOff))
plt.ylabel('number of people')
plt.xlabel('days')
plt.show()
import analysis
# A list of irreducible polynomials and constants
# http://www.sciencedirect.com/science/article/pii/S2212017313006051
irreducibles = [
0x11B, 0x11D, 0x12B, 0x12D, 0x139, 0x13F, 0x14D, 0x15F, 0x163, 0x165,
0x169, 0x171, 0x177, 0x17B, 0x187, 0x18B, 0x18D, 0x19F, 0x1A3, 0x1A9,
0x1B1, 0x1BD, 0x1C3, 0x1CF, 0x1D7, 0x1DD, 0x1E7, 0x1F3, 0x1F5, 0x1F9
]
constants = [
0x0A, 0x0F, 0x15, 0x2A, 0x2B, 0x31, 0x32, 0x35, 0x38, 0x40, 0x4A, 0x4E,
0x54, 0x5E, 0x62, 0x6E, 0x74, 0x7E, 0xF5, 0xF0, 0xEA, 0xD5, 0xD4, 0xCE,
0xCD, 0xCA, 0xC7, 0xBF, 0xB5, 0xB1, 0xAB, 0xA1, 0x9D, 0x91, 0x2B, 0x81
]
for p in irreducibles:
reduced = core._B2I(core._I2B(p)[1:])
for c in constants:
sbox = []
for elem in range(256):
sbox.append(
core.affine(core.inverse(elem, reduced), u=0x1F, v=0x63))
val = analysis.values(sbox)
out = "Polynomial: %02x, Constant: %02x, NL: %d, DP: %d"
print(out % (p, c, val[0], val[1]))
