Load Dependencies and Configuration Settings

In [1]:

import pandas as pd
import datetime
import math
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.mlab as mlab

%matplotlib inline

Load and View the Dataset

In [2]:

cs_df = pd.read_excel(io=r'Online Retail.xlsx')

In [3]:

cs_df.head()

Out[3]:

InvoiceNo StockCode Description Quantity InvoiceDate UnitPrice CustomerID Country

0 WHITE HANGING HEART T-LIGHT 2010-12-01 United

536365 85123A 6 2.55 17850.0
HOLDER 08:26:00 Kingdom

1 2010-12-01 United
536365 71053 WHITE METAL LANTERN 6 3.39 17850.0
08:26:00 Kingdom

2 2010-12-01 United
536365 84406B CREAM CUPID HEARTS COAT HANGER 8 2.75 17850.0
08:26:00 Kingdom

3 KNITTED UNION FLAG HOT WATER 2010-12-01 United

536365 84029G 6 3.39 17850.0
BOTTLE 08:26:00 Kingdom

4 2010-12-01 United
536365 84029E RED WOOLLY HOTTIE WHITE HEART. 6 3.39 17850.0
08:26:00 Kingdom

Transactions size

In [4]:
cs_df.shape

Out[4]:

(541909, 8)

Top Sales by Country

In [5]:

cs_df.Country.value_counts().reset_index().head(n=10)

Out[5]:

index Country

0 United
495478
Kingdom

1 Germany 9495
1 Germany 9495
index Country
2 France 8557

3 EIRE 8196

4 Spain 2533

5 Netherlands 2371

6 Belgium 2069

7 Switzerland 2002

8 Portugal 1519

9 Australia 1259

Top Customers contributing to 10% of total Sales

Number of customers

In [6]:
cs_df.CustomerID.unique().shape

Out[6]:
(4373,)

In [7]:
(cs_df.CustomerID.value_counts()/sum(cs_df.CustomerID.value_counts())*100).head(n=13).cumsum()

Out[7]:

17841.0 1.962249
14911.0 3.413228
14096.0 4.673708
12748.0 5.814728
14606.0 6.498553
15311.0 7.110850
14646.0 7.623350
13089.0 8.079807
13263.0 8.492020
14298.0 8.895138
15039.0 9.265809
14156.0 9.614850
18118.0 9.930462
Name: CustomerID, dtype: float64

Analyzing Data Quality Issues

Number of unique items

In [8]:
cs_df.StockCode.unique().shape

Out[8]:
(4070,)

Description of items: We see that the descriptions are more then the stock code so there must be some stock code which have more
than one decription

In [9]:
cs_df.Description.unique().shape
Out[9]:

(4224,)

In [10]:

cs_df.dtypes

Out[10]:
InvoiceNo object
StockCode object
Description object
Quantity int64
InvoiceDate datetime64[ns]
UnitPrice float64
CustomerID float64
Country object
dtype: object

In [11]:
cat_des_df = cs_df.groupby(["StockCode","Description"]).count().reset_index()

Stockcode which have more than one description

In [12]:

cat_des_df.StockCode.value_counts()[cat_des_df.StockCode.value_counts()>1].reset_index().head()

Out[12]:

index StockCode

0 20713 8

1 23084 7

2 21830 6

3 85175 6

4 85172 5

Example of one such stockcode

In [14]:
cs_df[cs_df['StockCode'] == cat_des_df.StockCode.value_counts()[cat_des_df.StockCode.value_counts(
)>1]
.reset_index()['index'][5]]['Description'].unique()

Out[14]:
array(['JUMBO BAG VINTAGE CHRISTMAS ', 'came coded as 20713',
'wrongly coded 20713', '20713 wrongly marked', 20713], dtype=object)

In [15]:
cs_df['invdatetime'] = pd.to_datetime(cs_df.InvoiceDate)

In [16]:
cs_df.Quantity.describe()

Out[16]:

count 541909.000000
mean 9.552250
std 218.081158
std 218.081158
min -80995.000000
25% 1.000000
50% 3.000000
75% 10.000000
max 80995.000000
Name: Quantity, dtype: float64

In [17]:

cs_df.UnitPrice.describe()

Out[17]:
count 541909.000000
mean 4.611114
std 96.759853
min -11062.060000
25% 1.250000
50% 2.080000
75% 4.130000
max 38970.000000
Name: UnitPrice, dtype: float64

Data Cleaning
In [18]:
# Seperate data for one geography
cs_df = cs_df[cs_df.Country == 'United Kingdom']

# Seperate attribute for total amount

cs_df['amount'] = cs_df.Quantity*cs_df.UnitPrice

# Remove negative or return transactions

cs_df = cs_df[~(cs_df.amount<0)]
cs_df.head()
cs_df = cs_df[~(cs_df.CustomerID.isnull())]

In [19]:
cs_df.shape

Out[19]:
(354345, 10)

Build Recency Feature

In [20]:
cs_df.InvoiceDate.max()

Out[20]:
Timestamp('2011-12-09 12:49:00')

In [21]:
cs_df.InvoiceDate.min()

Out[21]:
Timestamp('2010-12-01 08:26:00')

In [22]:
refrence_date = cs_df.InvoiceDate.max()
refrence_date = cs_df.InvoiceDate.max()
refrence_date = refrence_date + datetime.timedelta(days = 1)

In [23]:
cs_df['days_since_last_purchase'] = refrence_date - cs_df.InvoiceDate
cs_df['days_since_last_purchase_num'] = cs_df['days_since_last_purchase'].astype('timedelta64[D]')

Time period of transactions

In [24]:
customer_history_df = cs_df.groupby("CustomerID").min().reset_index()[['CustomerID',
'days_since_last_purchase_num']]
customer_history_df.rename(columns={'days_since_last_purchase_num':'recency'}, inplace=True)
customer_history_df.recency.describe()

Out[24]:

count 3921.000000
mean 92.188472
std 99.528995
min 1.000000
25% 18.000000
50% 51.000000
75% 143.000000
max 374.000000
Name: recency, dtype: float64

In [25]:

customer_history_df.head()

Out[25]:

CustomerID recency

0 12346.0 326.0

1 12747.0 2.0

2 12748.0 1.0

3 12749.0 4.0

4 12820.0 3.0

In [26]:
customer_history_df.recency.describe()

Out[26]:
count 3921.000000
mean 92.188472
std 99.528995
min 1.000000
25% 18.000000
50% 51.000000
75% 143.000000
max 374.000000
Name: recency, dtype: float64

In [27]:
x = customer_history_df.recency
mu = np.mean(customer_history_df.recency)
sigma = math.sqrt(np.var(customer_history_df.recency))
n, bins, patches = plt.hist(x, 1000, facecolor='green', alpha=0.75)
# add a 'best fit' line
y = mlab.normpdf(bins, mu, sigma)
l = plt.plot(bins, y, 'r--', linewidth=2)
l = plt.plot(bins, y, 'r--', linewidth=2)
plt.xlabel('Recency in days')
plt.ylabel('Number of transactions')
plt.title(r'$\mathrm{Histogram\ of\ sales\ recency}\ $')
plt.grid(True)

Build Frequency & Monetary value Features

In [28]:
customer_monetary_val = cs_df[['CustomerID', 'amount']].groupby("CustomerID").sum().reset_index()
customer_history_df = customer_history_df.merge(customer_monetary_val, how='outer')
customer_history_df.amount = customer_history_df.amount+0.001
customer_freq = cs_df[['CustomerID', 'amount']].groupby("CustomerID").count().reset_index()
customer_freq.rename(columns={'amount':'frequency'},inplace=True)
customer_history_df = customer_history_df.merge(customer_freq, how='outer')

Remove returns so that we only have purchases of a customer

In [29]:

customer_history_df.head()

Out[29]:

CustomerID recency amount frequency

0 12346.0 326.0 77183.601 1

1 12747.0 2.0 4196.011 103

2 12748.0 1.0 33719.731 4596

3 12749.0 4.0 4090.881 199

4 12820.0 3.0 942.341 59

In [30]:

from sklearn import preprocessing

import math

customer_history_df['recency_log'] = customer_history_df['recency'].apply(math.log)
customer_history_df['frequency_log'] = customer_history_df['frequency'].apply(math.log)
customer_history_df['amount_log'] = customer_history_df['amount'].apply(math.log)
feature_vector = ['amount_log', 'recency_log','frequency_log']
X_subset = customer_history_df[feature_vector].as_matrix()
scaler = preprocessing.StandardScaler().fit(X_subset)
X_scaled = scaler.transform(X_subset)

Visualizing Recency vs Monetary Value (scaled)

In [31]:
plt.scatter(customer_history_df.recency_log, customer_history_df.amount_log, alpha=0.5)

Out[31]:
<matplotlib.collections.PathCollection at 0x23fd2d19a90>

Visualizing Monetary Value distribution (scaled)

In [32]:
x = customer_history_df.amount_log
n, bins, patches = plt.hist(x, 1000, facecolor='green', alpha=0.75)

plt.xlabel('Log of Sales Amount')

plt.ylabel('Probability')
plt.title(r'$\mathrm{Histogram\ of\ Log\ transformed\ Customer\ Monetary\ value}\ $')
plt.grid(True)
#plt.show()

In [33]:

customer_history_df.head()

Out[33]:

CustomerID recency amount frequency recency_log frequency_log amount_log

0 12346.0 326.0 77183.601 1 5.786897 0.000000 11.253942

1 12747.0 2.0 4196.011 103 0.693147 4.634729 8.341890

2 12748.0 1.0 33719.731 4596 0.000000 8.432942 10.425838

3 12749.0 4.0 4090.881 199 1.386294 5.293305 8.316516

4 12820.0 3.0 942.341 59 1.098612 4.077537 6.848367

In [34]:

from mpl_toolkits.mplot3d import Axes3D

fig = plt.figure(figsize=(8, 6))

ax = fig.add_subplot(111, projection='3d')

xs =customer_history_df.recency_log
ys = customer_history_df.frequency_log
zs = customer_history_df.amount_log
ax.scatter(xs, ys, zs, s=5)

ax.set_xlabel('Recency')
ax.set_ylabel('Frequency')
ax.set_zlabel('Monetary')

#plt.show()

Out[34]:

<matplotlib.text.Text at 0x23fc2e56828>

Analyze Customer Segments with Clustering

In [34]:
from sklearn.cluster import KMeans
from sklearn.metrics import silhouette_samples, silhouette_score
import matplotlib.cm as cm

X = X_scaled

cluster_centers = dict()

for n_clusters in range(3,6,2):

fig, (ax1, ax2) = plt.subplots(1, 2)
#ax2 = plt.subplot(111, projection='3d')
fig.set_size_inches(18, 7)
ax1.set_xlim([-0.1, 1])
ax1.set_ylim([0, len(X) + (n_clusters + 1) * 10])

clusterer = KMeans(n_clusters=n_clusters, random_state=10)

cluster_labels = clusterer.fit_predict(X)

silhouette_avg = silhouette_score(X, cluster_labels)

cluster_centers.update({n_clusters :{
'cluster_center':clusterer.cluster_centers_,
'silhouette_score':silhouette_avg,
'labels':cluster_labels}
})

sample_silhouette_values = silhouette_samples(X, cluster_labels)

sample_silhouette_values = silhouette_samples(X, cluster_labels)
y_lower = 10
for i in range(n_clusters):
ith_cluster_silhouette_values = \
sample_silhouette_values[cluster_labels == i]

ith_cluster_silhouette_values.sort()

size_cluster_i = ith_cluster_silhouette_values.shape[0]
y_upper = y_lower + size_cluster_i

color = cm.spectral(float(i) / n_clusters)

ax1.fill_betweenx(np.arange(y_lower, y_upper),
0, ith_cluster_silhouette_values,
facecolor=color, edgecolor=color, alpha=0.7)

ax1.text(-0.05, y_lower + 0.5 * size_cluster_i, str(i))

y_lower = y_upper + 10 # 10 for the 0 samples

ax1.set_title("The silhouette plot for the various clusters.")

ax1.set_xlabel("The silhouette coefficient values")
ax1.set_ylabel("Cluster label")
ax1.axvline(x=silhouette_avg, color="red", linestyle="--")
ax1.set_yticks([])
ax1.set_xticks([-0.1, 0, 0.2, 0.4, 0.6, 0.8, 1])
colors = cm.spectral(cluster_labels.astype(float) / n_clusters)
feature1 = 0
feature2 = 2
ax2.scatter(X[:, feature1], X[:, feature2], marker='.', s=30, lw=0, alpha=0.7,
c=colors, edgecolor='k')

centers = clusterer.cluster_centers_
ax2.scatter(centers[:, feature1], centers[:, feature2], marker='o',
c="white", alpha=1, s=200, edgecolor='k')
for i, c in enumerate(centers):
ax2.scatter(c[feature1], c[feature2], marker='$%d$' % i, alpha=1,
s=50, edgecolor='k')
ax2.set_title("The visualization of the clustered data.")
ax2.set_xlabel("Feature space for the 1st feature i.e. monetary value")
ax2.set_ylabel("Feature space for the 2nd feature i.e. frequency")
plt.suptitle(("Silhouette analysis for KMeans clustering on sample data "
"with n_clusters = %d" % n_clusters),
fontsize=14, fontweight='bold')
#plt.show()
In [35]:

for i in range(3,6,2):
print("for {} number of clusters".format(i))
cent_transformed = scaler.inverse_transform(cluster_centers[i]['cluster_center'])
print(pd.DataFrame(np.exp(cent_transformed),columns=feature_vector))
print("Silhouette score for cluster {} is {}". format(i, cluster_centers[i]['silhouette_score']
))
print()

for 3 number of clusters

amount_log recency_log frequency_log
0 843.937271 44.083222 53.920633
1 221.236034 121.766072 10.668661
2 3159.294272 7.196647 177.789098
Silhouette score for cluster 3 is 0.30437444714898737

for 5 number of clusters

amount_log recency_log frequency_log
0 3905.544371 5.627973 214.465989
1 1502.519606 46.880212 92.306262
2 142.867249 126.546751 5.147370
3 408.235418 139.056216 25.530424
4 464.371885 13.386419 29.581298
Silhouette score for cluster 5 is 0.27958641427323727

Assign Cluster Labels

In [36]:

labels = cluster_centers[5]['labels']
customer_history_df['num_cluster5_labels'] = labels
labels = cluster_centers[3]['labels']
customer_history_df['num_cluster3_labels'] = labels

In [37]:
customer_history_df.head()

Out[37]:

CustomerID recency amount frequency recency_log frequency_log amount_log num_cluster5_labels num_cluster3_labels

0 12346.0 326.0 77183.601 1 5.786897 0.000000 11.253942 1 0

1 12747.0 2.0 4196.011 103 0.693147 4.634729 8.341890 0 2

2 12748.0 1.0 33719.731 4596 0.000000 8.432942 10.425838 0 2

3 12749.0 4.0 4090.881 199 1.386294 5.293305 8.316516 0 2

4 12820.0 3.0 942.341 59 1.098612 4.077537 6.848367 4 2

Visualize Segments
Visualize Segments
In [38]:

import plotly as py
import plotly.graph_objs as go
py.offline.init_notebook_mode()

x_data = ['Cluster 1','Cluster 2','Cluster 3','Cluster 4', 'Cluster 5']

cutoff_quantile = 100
field_to_plot = 'recency'

y0 = customer_history_df[customer_history_df['num_cluster5_labels']==0][field_to_plot].values
y0 = y0[y0<np.percentile(y0, cutoff_quantile)]
y1 = customer_history_df[customer_history_df['num_cluster5_labels']==1][field_to_plot].values
y1 = y1[y1<np.percentile(y1, cutoff_quantile)]
y2 = customer_history_df[customer_history_df['num_cluster5_labels']==2][field_to_plot].values
y2 = y2[y2<np.percentile(y2, cutoff_quantile)]
y3 = customer_history_df[customer_history_df['num_cluster5_labels']==3][field_to_plot].values
y3 = y3[y3<np.percentile(y3, cutoff_quantile)]
y4 = customer_history_df[customer_history_df['num_cluster5_labels']==4][field_to_plot].values
y4 = y4[y4<np.percentile(y4, cutoff_quantile)]
y_data = [y0,y1,y2,y3,y4]

colors = ['rgba(93, 164, 214, 0.5)', 'rgba(255, 144, 14, 0.5)', 'rgba(44, 160, 101, 0.5)',
'rgba(255, 65, 54, 0.5)', 'rgba(207, 114, 255, 0.5)', 'rgba(127, 96, 0, 0.5)']
traces = []

for xd, yd, cls in zip(x_data, y_data, colors):

traces.append(go.Box(
y=yd,
name=xd,
boxpoints=False,
jitter=0.5,
whiskerwidth=0.2,
fillcolor=cls,
marker=dict(
size=2,
),
line=dict(width=1),
))

layout = go.Layout(
title='Difference in sales {} from cluster to cluster'.format(field_to_plot),
yaxis=dict(
autorange=True,
showgrid=True,
zeroline=True,
dtick=50,
gridcolor='black',
gridwidth=0.1,
zerolinecolor='rgb(255, 255, 255)',
zerolinewidth=2,
),
margin=dict(
l=40,
r=30,
b=80,
t=100,
),
paper_bgcolor='white',
plot_bgcolor='white',
showlegend=False
)

fig = go.Figure(data=traces, layout=layout)

py.offline.iplot(fig)
In [39]:
x_data = ['Cluster 1','Cluster 2','Cluster 3','Cluster 4', 'Cluster 5']
cutoff_quantile = 80
field_to_plot = 'amount'
y0 = customer_history_df[customer_history_df['num_cluster5_labels']==0][field_to_plot].values
y0 = y0[y0<np.percentile(y0, cutoff_quantile)]
y1 = customer_history_df[customer_history_df['num_cluster5_labels']==1][field_to_plot].values
y1 = y1[y1<np.percentile(y1, cutoff_quantile)]
y2 = customer_history_df[customer_history_df['num_cluster5_labels']==2][field_to_plot].values
y2 = y2[y2<np.percentile(y2, cutoff_quantile)]
y3 = customer_history_df[customer_history_df['num_cluster5_labels']==3][field_to_plot].values
y3 = y3[y3<np.percentile(y3, cutoff_quantile)]
y4 = customer_history_df[customer_history_df['num_cluster5_labels']==4][field_to_plot].values
y4 = y4[y4<np.percentile(y4, cutoff_quantile)]
y_data = [y0,y1,y2,y3,y4]

colors = ['rgba(93, 164, 214, 0.5)', 'rgba(255, 144, 14, 0.5)', 'rgba(44, 160, 101, 0.5)',
'rgba(255, 65, 54, 0.5)', 'rgba(207, 114, 255, 0.5)', 'rgba(127, 96, 0, 0.5)']
traces = []

for xd, yd, cls in zip(x_data, y_data, colors):

traces.append(go.Box(
y=yd,
name=xd,
boxpoints=False,
jitter=0.5,
whiskerwidth=0.2,
fillcolor=cls,
marker=dict(
size=2,
),
line=dict(width=1),
))

layout = go.Layout(
title='Difference in sales {} from cluster to cluster'.format(field_to_plot),
yaxis=dict(
autorange=True,
showgrid=True,
zeroline=True,
dtick=1000,
gridcolor='black',
gridwidth=0.1,
zerolinecolor='rgb(255, 255, 255)',
zerolinewidth=2,
),
margin=dict(
l=40,
r=30,
b=80,
t=100,
),
paper_bgcolor='white',
 paper_bgcolor='white',
plot_bgcolor='white',
showlegend=False
)

fig = go.Figure(data=traces, layout=layout)

py.offline.iplot(fig)

In [40]:

x_data = ['Cluster 1','Cluster 2','Cluster 3','Cluster 4', 'Cluster 5']

cutoff_quantile = 80
field_to_plot = 'frequency'
y0 = customer_history_df[customer_history_df['num_cluster5_labels']==0][field_to_plot].values
y0 = y0[y0<np.percentile(y0, cutoff_quantile)]
y1 = customer_history_df[customer_history_df['num_cluster5_labels']==1][field_to_plot].values
y1 = y1[y1<np.percentile(y1, cutoff_quantile)]
y2 = customer_history_df[customer_history_df['num_cluster5_labels']==2][field_to_plot].values
y2 = y2[y2<np.percentile(y2, cutoff_quantile)]
y3 = customer_history_df[customer_history_df['num_cluster5_labels']==3][field_to_plot].values
y3 = y3[y3<np.percentile(y3, cutoff_quantile)]
y4 = customer_history_df[customer_history_df['num_cluster5_labels']==4][field_to_plot].values
y4 = y4[y4<np.percentile(y4, cutoff_quantile)]
y_data = [y0,y1,y2,y3,y4]

colors = ['rgba(93, 164, 214, 0.5)', 'rgba(255, 144, 14, 0.5)', 'rgba(44, 160, 101, 0.5)',
'rgba(255, 65, 54, 0.5)', 'rgba(207, 114, 255, 0.5)', 'rgba(127, 96, 0, 0.5)']
traces = []

for xd, yd, cls in zip(x_data, y_data, colors):

traces.append(go.Box(
y=yd,
name=xd,
boxpoints=False,
jitter=0.5,
whiskerwidth=0.2,
fillcolor=cls,
marker=dict(
size=2,
),
line=dict(width=1),
))

layout = go.Layout(
title='Difference in sales {} from cluster to cluster'.format(field_to_plot),
 title='Difference in sales {} from cluster to cluster'.format(field_to_plot),
yaxis=dict(
autorange=True,
showgrid=True,
zeroline=True,
dtick=100,
gridcolor='black',
gridwidth=0.1,
zerolinecolor='rgb(255, 255, 255)',
zerolinewidth=2,
),
margin=dict(
l=40,
r=30,
b=80,
t=100,
),
paper_bgcolor='white',
plot_bgcolor='white',
showlegend=False
)

fig = go.Figure(data=traces, layout=layout)

py.offline.iplot(fig)

In [41]:
x_data = ['Cluster 1','Cluster 2','Cluster 3']
cutoff_quantile = 100
field_to_plot = 'recency'
y0 = customer_history_df[customer_history_df['num_cluster3_labels']==0][field_to_plot].values
y0 = y0[y0<np.percentile(y0, cutoff_quantile)]
y1 = customer_history_df[customer_history_df['num_cluster3_labels']==1][field_to_plot].values
y1 = y1[y1<np.percentile(y1, cutoff_quantile)]
y2 = customer_history_df[customer_history_df['num_cluster3_labels']==2][field_to_plot].values
y2 = y2[y2<np.percentile(y2, cutoff_quantile)]

y_data = [y0,y1,y2]

colors = ['rgba(93, 164, 214, 0.5)', 'rgba(255, 144, 14, 0.5)', 'rgba(44, 160, 101, 0.5)',
'rgba(255, 65, 54, 0.5)', 'rgba(207, 114, 255, 0.5)', 'rgba(127, 96, 0, 0.5)']
traces = []

for xd, yd, cls in zip(x_data, y_data, colors):

traces.append(go.Box(
y=yd,
 y=yd,
name=xd,
boxpoints=False,
jitter=0.5,
whiskerwidth=0.2,
fillcolor=cls,
marker=dict(
size=2,
),
line=dict(width=1),
))

layout = go.Layout(
title='Difference in sales {} from cluster to cluster'.format(field_to_plot),
yaxis=dict(
autorange=True,
showgrid=True,
zeroline=True,
dtick=50,
gridcolor='black',
gridwidth=0.1,
zerolinecolor='rgb(255, 255, 255)',
zerolinewidth=2,
),
margin=dict(
l=40,
r=30,
b=80,
t=100,
),
plot_bgcolor='white',
showlegend=False
)

fig = go.Figure(data=traces, layout=layout)

py.offline.iplot(fig)

In [42]:
x_data = ['Cluster 1','Cluster 2','Cluster 3']
cutoff_quantile = 80
field_to_plot = 'amount'
y0 = customer_history_df[customer_history_df['num_cluster3_labels']==0][field_to_plot].values
y0 = y0[y0<np.percentile(y0, cutoff_quantile)]
y1 = customer_history_df[customer_history_df['num_cluster3_labels']==1][field_to_plot].values
y1 = customer_history_df[customer_history_df['num_cluster3_labels']==1][field_to_plot].values
y1 = y1[y1<np.percentile(y1, cutoff_quantile)]
y2 = customer_history_df[customer_history_df['num_cluster3_labels']==2][field_to_plot].values
y2 = y2[y2<np.percentile(y2, cutoff_quantile)]

y_data = [y0,y1,y2]

colors = ['rgba(93, 164, 214, 0.5)', 'rgba(255, 144, 14, 0.5)', 'rgba(44, 160, 101, 0.5)',
'rgba(255, 65, 54, 0.5)', 'rgba(207, 114, 255, 0.5)', 'rgba(127, 96, 0, 0.5)']
traces = []

for xd, yd, cls in zip(x_data, y_data, colors):

traces.append(go.Box(
y=yd,
name=xd,
boxpoints=False,
jitter=0.5,
whiskerwidth=0.2,
fillcolor=cls,
marker=dict(
size=2,
),
line=dict(width=1),
))

layout = go.Layout(
title='Difference in sales {} from cluster to cluster'.format(field_to_plot),
yaxis=dict(
dtick=1000,
)
)

fig = go.Figure(data=traces, layout=layout)

py.offline.iplot(fig)

In [43]:
x_data = ['Cluster 1','Cluster 2','Cluster 3']
cutoff_quantile = 90
field_to_plot = 'frequency'
y0 = customer_history_df[customer_history_df['num_cluster3_labels']==0][field_to_plot].values
y0 = y0[y0<np.percentile(y0, cutoff_quantile)]
y1 = customer_history_df[customer_history_df['num_cluster3_labels']==1][field_to_plot].values
y1 = y1[y1<np.percentile(y1, cutoff_quantile)]
y2 = customer_history_df[customer_history_df['num_cluster3_labels']==2][field_to_plot].values
y2 = y2[y2<np.percentile(y2, cutoff_quantile)]
y2 = y2[y2<np.percentile(y2, cutoff_quantile)]

y_data = [y0,y1,y2]

colors = ['rgba(93, 164, 214, 0.5)', 'rgba(255, 144, 14, 0.5)', 'rgba(44, 160, 101, 0.5)',
'rgba(255, 65, 54, 0.5)', 'rgba(207, 114, 255, 0.5)', 'rgba(127, 96, 0, 0.5)']
traces = []

for xd, yd, cls in zip(x_data, y_data, colors):

traces.append(go.Box(
y=yd,
name=xd,
boxpoints=False,
jitter=0.5,
whiskerwidth=0.2,
fillcolor=cls,
marker=dict(
size=2,
),
line=dict(width=1),
))

layout = go.Layout(
title='Difference in sales {} from cluster to cluster'.format(field_to_plot),
yaxis=dict(
autorange=True,
showgrid=True,
zeroline=True,
dtick=100,
gridcolor='black',
gridwidth=0.1,
zerolinecolor='rgb(255, 255, 255)',
zerolinewidth=2,
),
margin=dict(
l=40,
r=30,
b=80,
t=100,
),
paper_bgcolor='white',
plot_bgcolor='white',
showlegend=False
)

fig = go.Figure(data=traces, layout=layout)

py.offline.iplot(fig)