Data Analysis for Business Applications using Python | Desklib

This paper explores the use of data analysis and mining techniques to solve business problems faced by Infinity ecommerce company. It covers classification techniques, descriptive statistics, and predictive modelling for sales patterns and factors affecting sales. The study recommends better business practices for an edge over competitors. The data was obtained from the sales and financial records department for the past three years. The paper assumes that the products and factors affecting sales are uniformly distributed across all trade regions. Course code and college/university not mentioned.

Title:Data Analytics
Task: Data analysis using Python fro business application
Student’ Name:
Tutor’s Name:
Module Title:
Module Number:
Department:

Executive summary
Objective of paper
The purpose of our study is to employ different data mining and and data analysis
techniques in order to solve a number of business problems facing Infinity e-
commerce company.
What the paper covers
Our study explores classification techniques, descriptive statistics, and predictive
modelling for the data provided by the company in aim of determining the previous
sales patterns recorded by the sales and finance department, the factors that affect the
sales and how the products sold by the company are correlated. Further, we will
recommend better business practices that if employed will ensure an edge over our
competitors.
Assumptions
We assume that the products and the factors affecting the sales are uniformly
distributed across all the trade regions. We also assume that application of the
proposed methods will affect all the regions normally, i.e. all the variables whether
introduced or existent are normally distributed. In addition we assume that the cost of
advertisement reflects the extent of advertisement campaigns
Key words
Forecasting, Regression modelling, classification, SEM( Structural Equation
modelling)

Table Of Contents
Introduction..................................................................................................................4
Research Methodology.................................................................................................5
Analytical Results and Inference................................................................................8
Recommendations......................................................................................................15
Implementation plan for recommendations............................................................16
Conclusion...................................................................................................................17
Bibliography...............................................................................................................18
Appendix.....................................................................................................................19

Introduction
In the recent past there have been growing interest for the E-commerce sector of
business operations partly due to its non-geographical restrictions and also for its
security and ability to reach a wide range of customers despite their location.
Consequently, the interest has brewed competition between old players and new
entries into the business. Infinity company is an e-commerce company that
specialized in supply of Electronic products as well as other major products. The
company has a range of other products with reputable brands in clothes, household
commodities, toys, and gadgets. The electronic line of goods dealt with are:
 Samsung electronics
 Microsoft products (laptops and accessories)
 LG appliances
 Hp computer products and accessories
 Sony Home entertainment products
 AUCMA electronics
 Apple mobile and computer appliances
 Techno mobile phones and accessories
The recent concern of declining of sales expressed by the executive has bee attributed
to new competition as well as shift in consumer preference across the business
regions. The new development therefore prompts for new methods of business
approaches, even better, the innovation and adoption of suitable business methods
For our analysis we need to:
i. Examine sales patterns of the previous business years

ii. Examine the relationship between the sales and current business patterns
iii. Propose new methods of business sales promotion
To enable us explore all the aforementioned problems, we will use data classification
to explore relationships and regression to predict sales, given new business practices.
Research Methodology
Data
The data for this study was obtained from the sales department and the financial
records department for the past three years. The entries available from the data are as
in:
Variable description Size Denotation/ measure
Product 1386  Samsung electronics
 Microsoft products
(laptops and
accessories)
 LG appliances
 Hp computer products
and accessories
 Sony Home
entertainment
products
 AUCMA electronics
 Apple mobile and
computer appliances
 Techno mobile phones

and accessories
Shipping method 1386 Paid- P
Free- F
Sales recorded 1386 AUS dollars
Geographical region 1386 Asia
America
Europe
Australia
Other parts
Number of customers 1386 unspecified
Price of product 1386 AUS dollars
Customer type 1386 New- N
Existing- E
Advertisement 1386 AUS dollars
For our classification method we employed logistic regression and explored the
relationship between data variables. We use logistic regression in examining how
sales are influenced by shipping methods, also how sales are spread across the
marketing regions for the company. Moreover, we will determine the interrelationship
between the independent variables and the response variable. According to an article
on logistic regression by NCSS (2016), “Logistic regression analysis
studies the association between a categorical dependent variable and a set of
independent (explanatory) variables.” We used linear regression for predictive

modelling to predict the sales given different practices, for instance, how adoption of
product promotion methods such use of more sales persons and advertisement would
impact the sales recorded.
Use of forecasting in determining how different variables of interest are likely to play
out following a positive or negative variation of the variables against a response
variable will enable the company to near-accurately forecast on how the sales will be
made if we adopt different marketing methods and exportation of given good. For
instance through forecasting we determine how selling of related products together
are likely to affect total sales.

Analytical Results and Inference
General statistics of sales
Product Price (AUS
dollars)
Cost of
Advertisement
Customers Sales (AUS
dollars)
APPLE 1335055 1318993 6086 3054878.0
AUCMA 1614704 1566711 6839 3206451.0
HP 8580 7363 29 5172.0
Hp 1317170 1239529 6883 3570759.0
LG appliances 1462941 1483838 7050 3327564.0
Microsoft
electronics
1364818 1293901 6653 3281120.0
Samsung 1482765 1568058 7652 3531831.0
Sony 1386320 1386847 6648 3157285.0
Techno 224344 266384 848 0.0
From our statistics we deduce that the highest sales were recorded for Samsung
electronic appliances at 3531831.0 AUD while the least sales were by Techno mobile
at 0.0 AUD. This indicates that there was variance in the records made by the
different products. Additionally, the cost of advertisement incurred by the company
was such that promoting Samsung products had the highest revenue allocation while
HP and Techno had the least revenue allocated for advertisement.
The total sales for monthly sales preceding our analysis recorded were at AUD
23,135,060.

Analysis of regions
Export Region Price (AUS
dollars)
Cost of
Advertisement
Customers Sales (AUS
dollars)
Africa 2317687 2308667 10780 4921036.0
Australia 2015194 2002658 10372 4516548.0
Europe 1963479 1890778 9592 4787998.0
North America 3900337 3929521 17944 8909478.0
The highest sales were recorded from exports to North America at AUD 8909478.0
while the least were recorded from Australia at AUD 4516548.0, Africa was second
with sales of AUD 4921036.0 while Europe had sales of AUD 4787998.0. Also the
cost of advertisement in relation to regions were such that advertisement in North
America were the highest followed by Africa and Australia while the least
advertisements were done in Europe. Therefore for the business to ensure more sales,
it should export more to North America where we have the largest market.
Statistics of goods export to different regions
Product Export Region Price (AUS
dollars)
Cost of
Advertisement
Customers
APPLE Africa 331337 329349 1608
APPLE Australia 218470 218124 1100
APPLE Europe 129908 170510 607
APPLE North America 655340 601010 2771
AUCMA Africa 351519 320429 1420
AUCMA Australia 327375 337103 1489

AUCMA Europe 343520 296816 1610
AUCMA North America 592290 612363 2320
HP North America 8580 7363 29
Hp Africa 283226 246200 1370
Hp Australia 273359 277508 1498
Hp Europe 285316 303365 1616
Hp North America 475269 412456 2399
LG appliances Africa 310420 315892 1340
LG appliances Australia 324535 307948 1676
LG appliances Europe 309128 285657 1635
LG appliances North America 518858 574341 2399
Microsoft
electronics
Africa 311746 260592 1192
Microsoft
electronics
Australia 286825 287213 1664
Microsoft
electronics
Europe 347178 279859 1377
Microsoft
electronics
North America 419069 466237 2420
Samsung Africa 294825 365183 1914
Samsung Australia 268159 273902 1395
Samsung Europe 301610 289465 1596
Samsung North America 618171 639508 2747
Sony Africa 316442 336276 1539
Sony Australia 296485 279021 1403

Sony Europe 246819 265106 1151
Sony North America 526574 506444 2555
Techno Africa 118172 134746 397
Techno Australia 19986 21839 147
Techno North America 86186 109799 304
Product Export Region Price (AUS
dollars)
Cost of
Advertisement
Customers
From our analysis of the performance of different products in different regions we
found out that:
Price (AUS
dollars)
Cost of
Advertisement
Customers Sales (AUS
dollars)
count 32.0 32.0 32.0 32.0
mean 318646.78125 316613.25 1521.5 722970.625
std 154355.98657
83103
152344.73382
155778
709.37300802
61928
384421.32040
037913
min 8580.0 7363.0 29.0 0.0
25% 272059.0 263977.5 1303.0 591527.75
50% 309774.0 293140.5 1518.5 710760.0
75% 348263.25 344123.0 1735.5 843049.25
max 655340.0 639508.0 2771.0 1462076.0
The average cost of advertisement was 316613.25, the average number of customers
across the trade regions was 1521 while the averages of sales and total cost of the

products was 722970.625 and 318646.78125 respectively. Elsewhere the individual
products played across the regions as from table-4 above.
Product statistics
Product Customers Sales (AUS
dollars)
AUCMA 6839 3206451.0
Samsung 7652 3531831.0
LG appliances 7050 3327564.0
Sony 6648 3157285.0
Microsoft
electronics
6653 3281120.0
APPLE 6086 3054878.0
Hp 6883 3570759.0
Techno 848 0.0
HP 29 5172.0
Samsung had the most customers with the highest number of sales as well, techno had
848 customers with no sales recorded whereas HP had the least customers (29) with a
sales of 5172.0.
From our analysis the company exports more of Samsung products, it will be able to
increase its sales.
Effects of Advertisement
Price (AUS dollars) Customers Sales (AUS

dollars)
count 1131.0 1131.0 1131.0
mean 9015.64721485411 43.0486295313881
5
20455.4022988505
75
std 6840.10514113665
6
30.0123326476302
4
9701.73232838552
6
min 508.0 3.0 5152.0
25% 4222.5 21.0 12366.0
50% 7924.0 33.0 19812.0
75% 12197.0 63.0 28375.0
max 46813.0 203.0 70750.0
Advertisement had an effect of causing an average increase of AUD 20455.40
therefore there was a relationship between the sales and and the cost of advertisement.
The company can utilize advertisement as a means of promoting sales volumes
through increasing the advertisement budget.
Effect of free shipping on sales
shipping
Method
Price (AUS
dollars)
Cost of
Advertisement
Customers Sales (AUS
dollars)
Free 5150372 5250000 25323 11961382.0
Paid for 5046325 4881624 23365 11173678.0
Higher sales were recorded when free-shipping was used for the purchases made I.e.
11961382.0 compared to the sales for which shipping was paid for at 11173678.0.
From the analysis of shipping method, we conclude that for the company to ensure

return customers and also increase of sales, it should adopt measures such as free-
shipping and other product promotion incentives.
Regression of sales against geographical region
Using regression we find out that there is a relationship between the geographic
regions in which the company conducted business. For instance from earlier tables we
inference that North America and Africa comprise the company’s product market,
therefore the company should purpose to prioritize the two market spaces so as to
ensure more sales. Therefore, using forecasting we can view how varying the
company sales is more likely to play out owing to the current statistics and we can
project the probable sales through prioritizing company sales.

Recommendations
From the data analysis exercise we make the following recommendations for
consideration by the executive:
Regards to current business practices
i. The company should prioritize sales to North America and Africa mostly. As
well as considering employing of methods to curb competition from home firms
in Australia so as to increase sales. This can be through exploiting the avenues
through which the local consumers can make more purchases, including more
supply of popular electronics brands.
ii. The company should offer free shipping for every purchase made by both new
and existing customers, this will help foster a relationship and ensure more return
purchases.
iii. The company should drop redundant products such as Techno products and
concentrate on more sales generating products
New methods to increase sales
i. Given our analysis results, business methods such as more advertisement revenue
would ensure more use of new methods of advertisement such as celebrity
endorsement, free sampling, use of agents and more in order to increase product
awareness which will subsequently increase the sales volumes
ii. Also the company should try the affiliate sales program, where there are referrals
by interested independent marketers who gain commission through sales made by
their referrals. This would be an excellent method to get new customers to buy
our products due to the enormous potential of affiliate marketing

Implementation plan for recommendations
Below we recommend the best method through which the company can ensure that
the recommendations are productive after implementation if the executive adopts the
recommendations, then the sales department can anticipate more sales in the coming
financial year. The implementation plan includes:
i. Set the minimum export volumes to the regions at a value of USD500,000 with a
plan of varying them with a ratio of the scale 0.10 on the current ratio, such that
first the measure of the sales at the end of the coming financial year is recorded
and the executive determines whether the current sales increase projection ratio is
low or should be modified or maintained
ii. Adopt a mechanism to ensure the optimal revenue for advertisement purposes is
set aside such that we do not over advertise or under advertise, this will serve the
purpose of both checking the overall expenditure and ensure increased product
awareness
iii. The sales department should adopt free samples, affiliate marketing, and use of
coupons as a measure of advertising. Affiliate marketers can be found through
use of websites and referral advertisement by bloggers and interested parties.

Conclusion
In conclusion, we have found out that, as time changes methods of business
conducting change too, in a way we can say they are linearly related. In order to
ensure the firm’s sustainability, the executive ought to employ methods that ensure
the company evolves with time. One of the best measures against which the firm can
program itself to adapt with change is through data analysis. Data analysis provides
means through which the company can gain insights on the performance as well as
gauge its sustainability prospects through factors such as how much profits it
generates.
Even more interesting is the role of data analysis in business intelligence and decision
making. Generally, data is key to better decisions and eventual growth of business and
ought to be fully adopted.

Bibliography
Omar, H., Hoang, V.H. and Liu, D.R., 2016. A Hybrid Neural Network Model for
Sales Forecasting Based on ARIMA and Search Popularity of Article
Titles. Computational intelligence and neuroscience, 2016.
Donald H. & Tim S. (2008).Business leaders speak out: their real strategic problems..
Journal of Business Strategy, Vol. 29 Issue: 5, pp.32-37,
https://doi.org/10.1108/02756660810902305.
Adi, B. (2017). Simple and Multiple Linear Regression in Python. Journal of data
Science. Vol. 2, No. 12, pp 23-24. Accessed from: Logistic Regression using Python
(scikit-learn) – Towards Data Science.
Fernandes,C & Nicole, I. (2018) The influence of corporate social responsibility
associations on consumers’ perceptions towards global brands. Journal of Strategic
Marketing, pp 39-57, DOI: 10.1080/0965254X.2018.1464497.
Van Donselaar, K.H., Peters, J., De Jong, A. and Broekmeulen, R.A.C.M., 2016.
Analysis and forecasting of demand during promotions for perishable
items. International Journal of Production Economics, 172, pp.65-75.
Žylius, G.G., Simutis, R. and Vaitkus, V., 2015. Evaluation of computational
intelligence techniques for daily product sales forecasting. International Journal of
Computing, 14(3), pp.157-164.
Baardman, L., Levin, I., Perakis, G. and Singhvi, D., 2017. Leveraging Comparables
for New Product Sales Forecasting.

Appendix
# -*- coding: utf-8 -*-
"""
Created on Thu Jun 7 22:17:48 2018
@author: wyc
"""
# -*- coding: utf-8 -*-
"""
Created on Thu Jun 7 22:17:48 2018
@author: wyc
"""
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
datafrm=pd.read_csv("Infinitydata.csv")
datafrm.describe()
datafrm.head()
datafrm["Export Region"].unique()i
datafrm["Product"].unique()
def unique_counts(datafrm):
for i in datafrm.columns:
count = datafrm[i].nunique()
print(i, ": ", count)
unique_counts(datafrm)
from scipy.stats import trim_mean, kurtosis
trim_mean
kurtosis
from scipy.stats.mstats import mode, gmean, hmean
mode
gmean
hmean
desrpt=datafrm.describe()
from sklearn.linear_model import LinearRegression
LinearRegression
import statsmodels.api as sm
sm
#datafrm.groupby(["Product"]).sum().sort_values("Export Region", ascending=False)
group0 = datafrm.groupby(["Export Region"]).sum()
group0.head(50)
datagrp2 = datafrm.groupby(["shipping Method"]).sum()
datagrp2.head(50)
datagrp22 = datafrm.groupby(["Cost of Advertisement"]).sum()
datagrp22.head(50)
desrpt55=datagrp22.describe()
datagrp3 = datafrm.groupby(["Product","Export Region" ]).sum()
datagrp3.head(50)
datagrp1 = datafrm.groupby(["Product"]).sum()
datagrp1.head(50)
#mask=(datafrm['Export Region'].values =='Europe')
#mask
dta=datafrm.loc[datafrm['Export Region'] == 'North America']
dta
dta1=datafrm.loc[datafrm['Export Region'] == 'Africa']
dta2=datafrm.loc[datafrm['Export Region'] == 'Europe']
dta3=datafrm.loc[datafrm['Export Region'] == 'Africa']

dta4=datafrm.loc[datafrm['Export Region'] == 'Australia']
desrpt3=dta1.describe()
desrpt4=dta2.describe()
desrpt41=dta3.describe()
desrpt5=dta4.describe()
desrpt6=datagrp3.describe()
lf=dta.loc[dta['Product'] == 'Samsung']
classfeddata = lf.drop(['price','shipping Method','Customer type (N=New- customer)','Export
Region','Product'], axis=1)
classfeddata
lf=dta.loc[dta['Product'] == 'Samsung']
classfeddata = lf.drop(['price','shipping Method','Customer type (N=New- customer)','Export
Region','Product'], axis=1)
classfeddata
descrp=classfeddata.describe()
from sklearn.linear_model import LinearRegression
LinearRegression
from sklearn.cross_validation import train_test_split
train_test_split
import statsmodels.api as sm
sm
plt.style.use('seaborn')
plt=datafrm.plot(x='Sales (AUS dollars)', y='Export Region', kind='scatter')
plt.show()
lr = LinearRegression()
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
from sklearn import datasets, linear_model
def get_data(file_name):
x_parameter = []
y_parameter = []
for single_prices ,single_value in zip(dta['Sales (AUS dollars)'],dta['Customers']):
x_parameter.append([float(single_prices)])
y_parameter.append(float(single_value))
return x_parameter,y_parameter
x,y = get_data(dta)
print (x)
print(y)
# Function for Fitting data to Linear model
def linear_model_main(X_parameters,Y_parameters,predict_value):
# Create linear regression object
regr = linear_model.LinearRegression()
regr.fit(X_parameters, Y_parameters)
predict_outcome = regr.predict(predict_value)
predictions = {}
predictions['intercept'] = regr.intercept_
predictions['coefficient'] = regr.coef_
predictions['predicted_value'] = predict_outcome
return predictions
x,y = get_data('input_data.csv')
predict_value = 1000
result = linear_model_main(x,y,predict_value)
print ("Intercepting value " , result['intercept'])
print ("value of the coefficient" , result['coefficient'])
print ("Prediction value: ",result['predicted_value'])
# Function to show the resutls of linear fit model
def show_linear_line(X_parameters,Y_parameters):
# Create linear regression object
regr = linear_model.LinearRegression()
regr.fit(X_parameters, Y_parameters)
plt.scatter(X_parameters,Y_parameters,color='blue')
plt.plot(X_parameters,regr.predict(X_parameters),color='red',linewidth=4)
plt.xticks(())
plt.yticks(())
plt.show()