def plot_pv_fraction_in_campaign(pv, access_method, counts, sample_rate, hours = 1):
# get pvs per hours for source
pv = copy.copy(pv[pv['access_method'] == access_method])
pv.index = pd.Series(pv.index).apply(lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
pv = pv.groupby(pv.index).sum()
pv['pageviews_in_campaign'] = get_counts_df(counts, hours, sample_rate).sum(axis=1)
pv['fraction_of_pageviews_in_campaign'] = pv['pageviews_in_campaign'] / pv['pageviews']
plot_df(pv[['fraction_of_pageviews_in_campaign']], 'fraction per %d hours' % hours)
return pv
def plot_central_tendency(d,
regs,
start='2000',
stop='2050',
hours=1,
index=None,
ylabel='dollars per donation',
title='',
method='mean',
amount_limit=10000.0):
d = d[start:stop]
if index is None:
d.index = pd.Series(d.index).apply(
lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
else:
d.index = d[index]
d_plot = pd.DataFrame()
for name, reg in regs.items():
amounts = d.ix[d.name.str.match(reg).apply(bool)]['amount']
amounts[amounts > amount_limit] = amount_limit
if method == 'mean':
d_plot[name] = amounts.groupby(
amounts.index).sum() / amounts.groupby(amounts.index).count()
else:
d_plot[name] = amounts.groupby(amounts.index).median()
return plot_df(d_plot, ylabel=ylabel, title=title, interactive=False)
def plot_by_time(d, regs, start = '2000', stop = '2050', hours = 1, amount = False, cum = False, normalize = False, ylabel = '', interactive = False, index = None, rotate=False):
d = d[start:stop]
if index is None:
d.index = pd.Series(d.index).apply(lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
else:
d.index = d[index]
d_plot = pd.DataFrame()
for name, reg in regs.items():
if amount:
counts = d.ix[d.name.str.match(reg).apply(bool)]['amount']
else:
counts = d.ix[d.name.str.match(reg).apply(bool)]['n']
if normalize:
counts = counts/counts.sum()
if cum:
d_plot[name] = counts.groupby(counts.index).sum().cumsum()
else:
d_plot[name] = counts.groupby(counts.index).sum()
d_plot = d_plot.fillna(0)
#d_plot.plot(figsize=(10, 4))
return plot_df(d_plot, ylabel, interactive = interactive, rotate=rotate)
def plot_by_impressions_seen(d,
regs,
normalize=True,
max_impressions=5,
amount=False,
interactive=False,
ylabel=''):
d_plot = pd.DataFrame()
for name, reg in sorted(regs.items()):
if amount:
counts = d[d.name.str.contains(reg)]['amount']
else:
counts = d[d.name.str.contains(reg)]['n']
counts = counts.groupby(counts.index).sum()
if normalize:
counts = counts / counts.sum()
counts1 = counts.loc[:max_impressions]
counts1.loc[max_impressions + 1] = counts[max_impressions:].sum()
d_plot[name] = counts1
return plot_df(d_plot,
xlabel='impressions seen',
ylabel=ylabel,
interactive=interactive)
def plot_rate_by_time(don,
imp,
regs,
hours=1,
start='2000',
stop='2050',
ylabel='donation rate',
interactive=False,
index=None):
don = don[start:stop]
imp = imp[start:stop]
if index is None:
don.index = pd.Series(don.index).apply(
lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
imp.index = pd.Series(imp.index).apply(
lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
else:
don.index = don[index]
imp.index = imp[index]
d_plot = pd.DataFrame()
for name, reg in regs.items():
dons = don.ix[don.name.str.match(reg).apply(bool)]['n']
dons = dons.groupby(dons.index).sum()
imps = imp.ix[imp.name.str.match(reg).apply(bool)]['n']
imps = imps.groupby(imps.index).sum()
d_plot[name] = dons / imps
#d_plot.plot(figsize=(10, 4))
return plot_df(d_plot, ylabel, interactive=interactive)
def plot_central_tendency(
d,
regs,
start="2000",
stop="2050",
hours=1,
index=None,
ylabel="dollars per donation",
title="",
method="mean",
amount_limit=10000.0,
):
d = d[start:stop]
if index is None:
d.index = pd.Series(d.index).apply(lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
else:
d.index = d[index]
d_plot = pd.DataFrame()
for name, reg in regs.items():
amounts = d.ix[d.name.str.match(reg).apply(bool)]["amount"]
amounts[amounts > amount_limit] = amount_limit
if method == "mean":
d_plot[name] = amounts.groupby(amounts.index).sum() / amounts.groupby(amounts.index).count()
else:
d_plot[name] = amounts.groupby(amounts.index).median()
return plot_df(d_plot, ylabel=ylabel, title=title, interactive=False)
def plot_by_impressions_seen(d,
regs,
normalize = True,
max_impressions = 5,
amount = False,
interactive = False,
ylabel = ''
):
d_plot = pd.DataFrame()
for name, reg in sorted(regs.items()):
if amount:
counts = d[d.name.str.contains(reg)]['amount']
else:
counts = d[d.name.str.contains(reg)]['n']
counts = counts.groupby(counts.index).sum()
if normalize:
counts = counts / counts.sum()
counts1 = counts.loc[:max_impressions]
counts1.loc[max_impressions+1] = counts[max_impressions:].sum()
d_plot[name] = counts1
return plot_df(d_plot, xlabel = 'impressions seen', ylabel=ylabel, interactive = interactive)
def plot_conversion_rate(d, regs, start = '2000', stop = '2050', hours = 1, index = None, ylabel = 'conversion_rate',title= ''):
if d.shape[0] == 0:
print ('No Conversion rate data for this device')
return
d = d[start:stop]
if index is None:
d.index = pd.Series(d.index).apply(lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
else:
d.index = d[index]
d_plot = pd.DataFrame()
for name, reg in regs.items():
clicks = d.ix[d.name.str.match(reg).apply(bool)]
if clicks.shape[0] == 0:
continue
for method in methods:
clicks_by_method = clicks[clicks['payment_method'] == method]
if clicks_by_method.shape[0] == 0:
continue
donations = clicks_by_method[clicks_by_method['donation'] == 1]['n']
if donations.shape[0] == 0:
continue
donations = donations.groupby(donations.index).sum()
clicks_by_method = clicks_by_method.groupby(clicks_by_method.index)['n'].sum()
d_plot[name+' '+ method] = donations / clicks_by_method
if d_plot.shape[0] < 3:
print ('No Conversion rate data for this device')
return
return plot_df(d_plot, ylabel=ylabel, title=title, interactive = False)
def plot_donation_rate(don,
imp,
regs,
cum = False,
max_impressions=5,
amount = False,
interactive = False,
ylabel = '',
):
d_plot = pd.DataFrame()
for name, reg in sorted(regs.items()):
if amount:
donations = don.ix[don.name.str.contains(reg)]['amount']
else:
donations = don.ix[don.name.str.contains(reg)]['n']
donations = donations.groupby(donations.index).sum()
donations1 = donations.loc[:max_impressions]
donations1.loc[max_impressions+1] = donations[max_impressions:].sum()
impressions = imp.ix[imp.name.str.contains(reg)]['n']
impressions = impressions.groupby(impressions.index).sum()
impressions1 = impressions.loc[:max_impressions]
impressions1.loc[max_impressions+1] = impressions[max_impressions:].sum()
if cum:
impressions1 = impressions1.cumsum()
donations1 = donations1.cumsum()
d_plot[name] = donations1 / impressions1
return plot_df(d_plot, xlabel = 'impressions seen', ylabel=ylabel, interactive=interactive)
def plot_conversion_rate(d,
regs,
start='2000',
stop='2050',
hours=1,
index=None,
ylabel='conversion_rate',
title='',
methods=['amazon', 'paypal', 'cc']):
d = d[start:stop]
if index is None:
d.index = pd.Series(d.index).apply(
lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
else:
d.index = d[index]
d_plot = pd.DataFrame()
for name, reg in regs.items():
clicks = d.ix[d.name.str.match(reg).apply(bool)]
for method in methods:
clicks_by_method = clicks[clicks['payment_method'] == method]
donations = clicks_by_method[clicks_by_method['donation'] ==
1]['n']
donations = donations.groupby(donations.index).sum()
clicks_by_method = clicks_by_method.groupby(
clicks_by_method.index)['n'].sum()
#print clicks_by_method.head()
#print donations.head()
d_plot[name + ' ' + method] = donations / clicks_by_method
return plot_df(d_plot, ylabel=ylabel, title=title, interactive=False)
def plot_by_time(d,
regs,
start='2000',
stop='2050',
hours=1,
amount=False,
cum=False,
normalize=False,
ylabel='',
interactive=False,
index=None,
rotate=False):
d = d[start:stop]
if index is None:
d.index = pd.Series(d.index).apply(
lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
else:
d.index = d[index]
d_plot = pd.DataFrame()
for name, reg in regs.items():
if amount:
counts = d.ix[d.name.str.match(reg).apply(bool)]['dollars']
else:
counts = d.ix[d.name.str.match(reg).apply(bool)]['n']
if normalize:
counts = counts / counts.sum()
if cum:
d_plot[name] = counts.groupby(counts.index).sum().cumsum()
else:
d_plot[name] = counts.groupby(counts.index).sum()
if d_plot[name].shape[0] < 3:
del d_plot[name]
if d_plot.shape[0] < 3:
print('There is no data for this campaign or this kind of banners')
return
d_plot = d_plot.fillna(0)
plot_df(d_plot, ylabel, interactive=interactive, rotate=rotate)
def plot_pv_fraction_in_campaign(pv,
access_method,
counts,
sample_rate,
hours=1):
# get pvs per hours for source
pv = copy.copy(pv[pv['access_method'] == access_method])
pv.index = pd.Series(pv.index).apply(
lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
pv = pv.groupby(pv.index).sum()
pv['pageviews_in_campaign'] = get_counts_df(counts, hours,
sample_rate).sum(axis=1)
pv['fraction_of_pageviews_in_campaign'] = pv['pageviews_in_campaign'] / pv[
'pageviews']
plot_df(pv[['fraction_of_pageviews_in_campaign']],
'fraction per %d hours' % hours)
return pv
def plot_traffic(pv, imp, source, reg, start, stop, hours = 1):
pv_s = copy.copy(pv[pv['access_method'] == source])
pv_s.index = pd.Series(pv_s.index).apply(lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
pv_s = pv_s.groupby(pv_s.index).sum()
pv_s.rename(columns={'pageviews': source + ' pageviews'}, inplace=True)
imp_s = copy.copy(imp.ix[imp.name.str.match(reg).apply(bool)][['n']])
imp_s.index = pd.Series(imp_s.index).apply(lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
imp_s = imp_s.groupby(imp_s.index).sum()
imp_s.rename(columns={'n': source + ' impressions'}, inplace=True)
d = pv_s.merge(imp_s, how = 'left', left_index = True, right_index = True)[[source + ' pageviews', source + ' impressions']]
return plot_df(d, 'count per %d hours' % hours)
def plot_by_time(d, regs, start = '2000', stop = '2050', hours = 1, amount = False, cum = False, normalize = False, ylabel = '', interactive = False, index = None, rotate=False):
d = d[start:stop]
if index is None:
d.index = pd.Series(d.index).apply(lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
else:
d.index = d[index]
d_plot = pd.DataFrame()
for name, reg in regs.items():
if amount:
counts = d.ix[d.name.str.match(reg).apply(bool)]['dollars']
else:
counts = d.ix[d.name.str.match(reg).apply(bool)]['n']
if normalize:
counts = counts/counts.sum()
if cum:
d_plot[name] = counts.groupby(counts.index).sum().cumsum()
else:
d_plot[name] = counts.groupby(counts.index).sum()
if d_plot[name].shape[0] < 3:
del d_plot[name]
if d_plot.shape[0] < 3:
print('There is no data for this campaign or this kind of banners')
return
d_plot = d_plot.fillna(0)
plot_df(d_plot, ylabel, interactive = interactive, rotate=rotate)
def plot_traffic(pv, imp, source, reg, start, stop, hours=1):
pv_s = copy.copy(pv[pv['access_method'] == source])
pv_s.index = pd.Series(pv_s.index).apply(
lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
pv_s = pv_s.groupby(pv_s.index).sum()
pv_s.rename(columns={'pageviews': source + ' pageviews'}, inplace=True)
imp_s = copy.copy(imp.ix[imp.name.str.match(reg).apply(bool)][['n']])
imp_s.index = pd.Series(imp_s.index).apply(
lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
imp_s = imp_s.groupby(imp_s.index).sum()
imp_s.rename(columns={'n': source + ' impressions'}, inplace=True)
d = pv_s.merge(imp_s, how='left', left_index=True, right_index=True)[[
source + ' pageviews', source + ' impressions'
]]
return plot_df(d, 'count per %d hours' % hours)
def plot_rate_by_time(don,
imp,
regs,
hours=1,
start='2000',
stop='2050',
ylabel='donation rate',
interactive=False,
index=None):
don = don[start:stop]
imp = imp[start:stop]
if index is None:
don.index = pd.Series(don.index).apply(
lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
imp.index = pd.Series(imp.index).apply(
lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
else:
don.index = don[index]
imp.index = imp[index]
d_plot = pd.DataFrame()
for name, reg in regs.items():
dons = don.ix[don.name.str.match(reg).apply(bool)]['n']
dons = dons.groupby(dons.index).sum()
imps = imp.ix[imp.name.str.match(reg).apply(bool)]['n']
imps = imps.groupby(imps.index).sum()
d_rate = dons / imps
if d_rate.shape[0] < 3:
continue
largest = d_rate.nlargest(2).ix[1]
d_rate[d_rate > largest] = largest
d_plot[name] = d_rate
if d_plot.shape[0] < 3:
print('There is no data for this campaign or this kind of banners')
return
return plot_df(d_plot, ylabel, interactive=interactive)
def plot_conversion_rate(d, regs, start = '2000', stop = '2050', hours = 1, index = None, ylabel = 'conversion_rate',title= '', methods = ['amazon', 'paypal', 'cc']):
d = d[start:stop]
if index is None:
d.index = pd.Series(d.index).apply(lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
else:
d.index = d[index]
d_plot = pd.DataFrame()
for name, reg in regs.items():
clicks = d.ix[d.name.str.match(reg).apply(bool)]
for method in methods:
clicks_by_method = clicks[clicks['payment_method'] == method]
donations = clicks_by_method[clicks_by_method['donation'] == 1]['n']
donations = donations.groupby(donations.index).sum()
clicks_by_method = clicks_by_method.groupby(clicks_by_method.index)['n'].sum()
#print clicks_by_method.head()
#print donations.head()
d_plot[name+' '+ method] = donations / clicks_by_method
return plot_df(d_plot, ylabel=ylabel, title=title, interactive = False)
def plot_donation_rate(
don,
imp,
regs,
cum=False,
max_impressions=5,
amount=False,
interactive=False,
ylabel='',
):
d_plot = pd.DataFrame()
for name, reg in sorted(regs.items()):
if amount:
donations = don.ix[don.name.str.contains(reg)]['amount']
else:
donations = don.ix[don.name.str.contains(reg)]['n']
donations = donations.groupby(donations.index).sum()
donations1 = donations.loc[:max_impressions]
donations1.loc[max_impressions + 1] = donations[max_impressions:].sum()
impressions = imp.ix[imp.name.str.contains(reg)]['n']
impressions = impressions.groupby(impressions.index).sum()
impressions1 = impressions.loc[:max_impressions]
impressions1.loc[max_impressions +
1] = impressions[max_impressions:].sum()
if cum:
impressions1 = impressions1.cumsum()
donations1 = donations1.cumsum()
d_plot[name] = donations1 / impressions1
return plot_df(d_plot,
xlabel='impressions seen',
ylabel=ylabel,
interactive=interactive)
def plot_by_time(d,
regs,
start='2000',
stop='2050',
hours=1,
amount=False,
cum=False,
normalize=False,
ylabel='',
interactive=False,
index=None,
rotate=False):
d = d[start:stop]
if index is None:
d.index = pd.Series(d.index).apply(
lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
else:
d.index = d[index]
d_plot = pd.DataFrame()
for name, reg in regs.items():
if amount:
counts = d.ix[d.name.str.match(reg).apply(bool)]['amount']
else:
counts = d.ix[d.name.str.match(reg).apply(bool)]['n']
if normalize:
counts = counts / counts.sum()
if cum:
d_plot[name] = counts.groupby(counts.index).sum().cumsum()
else:
d_plot[name] = counts.groupby(counts.index).sum()
d_plot = d_plot.fillna(0)
#d_plot.plot(figsize=(10, 4))
return plot_df(d_plot, ylabel, interactive=interactive, rotate=rotate)
def plot_rate_by_time(don, imp, regs, hours = 1, start = '2000', stop = '2050', ylabel = 'donation rate', interactive = False, index = None):
don = don[start:stop]
imp = imp[start:stop]
if index is None:
don.index = pd.Series(don.index).apply(lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
imp.index = pd.Series(imp.index).apply(lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
else:
don.index = don[index]
imp.index = imp[index]
d_plot = pd.DataFrame()
for name, reg in regs.items():
dons = don.ix[don.name.str.match(reg).apply(bool)]['n']
dons = dons.groupby(dons.index).sum()
imps = imp.ix[imp.name.str.match(reg).apply(bool)]['n']
imps = imps.groupby(imps.index).sum()
d_plot[name] = dons/imps
#d_plot.plot(figsize=(10, 4))
return plot_df(d_plot, ylabel, interactive = interactive)
def plot_rate_by_time(don, imp, regs, hours = 1, start = '2000', stop = '2050', ylabel = 'donation rate', interactive = False, index = None):
don = don[start:stop]
imp = imp[start:stop]
if index is None:
don.index = pd.Series(don.index).apply(lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
imp.index = pd.Series(imp.index).apply(lambda tm: tm - timedelta(hours=(24 * tm.day + tm.hour) % hours))
else:
don.index = don[index]
imp.index = imp[index]
d_plot = pd.DataFrame()
for name, reg in regs.items():
dons = don.ix[don.name.str.match(reg).apply(bool)]['n']
dons = dons.groupby(dons.index).sum()
imps = imp.ix[imp.name.str.match(reg).apply(bool)]['n']
imps = imps.groupby(imps.index).sum()
d_rate = dons/imps
if d_rate.shape[0] < 3:
continue
largest = d_rate.nlargest(2).ix[1]
d_rate[d_rate > largest] = largest
d_plot[name] = d_rate
if d_plot.shape[0] < 3:
print('There is no data for this campaign or this kind of banners')
return
return plot_df(d_plot, ylabel, interactive = interactive)
