Abstract. and prior work in the science

 

Abstract. In
the Ultramodern age of technology, anticipation of market trend is very
important to observe consumer behaviour in this competitive world as trends are
volatile.Building on developments in machine learning and
prior work in the science of behaviour prediction, we construct a model
designed to predict the behavior of Consumer. The aim of this research paper is to examine the
relation between consumer behavior parameters and willingness to buy. First we investigate to find relationship between consumer behavior changing parameters (environmental factor,
organizational factor, Individual factor, Interpersonal factor) to consumer
willingness to buy products. These parameters affect
the choice of purchasing the product significantly. Second, we develop a time-evolving random forest
classifier that leverages unique feature engineering to predict the behavior
accurately. Traditional methods are not
much effective and cannot be considered for long term. These methods are not
able to address various problems while predicting consumer behavior. So, objective of this
research paper is to give appropriate solution to all such problems and to
reduce forecasting errors with the help of CLASSIFICATION
MODEL, terminology of machine learning. The Algorithm, which is used here
is providing more accuracy than other algorithms.

Key Words:
Random Forest Classifier, feature extraction.

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

 

1  
Introduction

Consuming is an
inevitable part of life. Purchasing depends on the choice and utility of the consumers. Therefore,
Consumers plays a vital role in purchasing. In purchasing a product  to know the consumers mind set is important . Some products appeal to customers while the same one does not appeal to some other. Hence,Buying
behaviors usually differs person to person.

 To keep up with changes, understanding a consumers requires current analytical methods that refine on data points to reveal the behavior. Analyzing customer
behavior is a very challenging task so it is very vital to know which strategy
should adopt.In the modern
technological world, there is need of Innovative marketing for analysing
consumer behavior.

 

So it is very important to understand why,
when, how and what other factors that influence buying decision of the consumers.For predicting consumer behavior
sagacious methods required other than finding out what
they had
purchased earlier. Consumers can classify according to their actual behaviors.

 

This process consists of five sections. Second section deals with  literature
surveys. It discuss the  factors influencing consumers behaviour, performance of
Forest Trees algorithm, comparison of different approaches, and drawbacks of the algorithm. Third section explains the
methodology to deduct the buying behavior of consumer using Random Forest Trees
algorithm. Implementation and results discussed in section 4 which find the best configurations for
higher prediction accuracy.

2      Literature
Review

There are several factors such as external and
internal factors which influence consumer their purchasing process and
decision. External factors such as cultural factor and social factor depend on
nationalities, geographic regions, racial groups, religions, income,
profession, and education whereas reference groups, family, role and status.
Internal
factors such as age, profession, education,, income,
personality and life style are depend on what to buy or what not to buy1. Sometimes it’s also depend on
the psychological factors such as perception, motivation, learning from past
experience and beliefs and attitude. Below given are some factors which
influences consumer buying behavior2.

2.1  
Factors influence consumer behavior

1.     
Individual: 
It refers to a demographic factors, sex, race, age etc for a particular
person.  Sometimes purchasing is also
depending on the decision maker in the family. It is observed that youth buy things for divergent
reasons than other age group people.

2.     
Environmental: Culture majorly affects the buying behaviour of every
customer and therefore the marketing experts should try to dissect the market
on cultural grounds and likes/dislikes/wants of customers. Our surroundings
represents a mixture of norms, convictions, attitudes, financial values, moral
values and habits developed in time by humans, which the members of our society
share and which also determine their behaviour, including the buying and
expending behaviour .It has two main effects on the market trend: it tells us
about the most fundamental values that affect the consumer behaviour plan, and it can also be used to figure out and identify the
subcultures that reflect on the meaningful marketing departments and
opportunities . Also, “a person's utilisation behaviour” can be
understood and copied or rejected by our peer group. In the end it results into
become the company's norm of behaviour and is recognized as an
essential part of the environment of a neighbourhood.

 

3.      Organizational: The internal factor affecting the
buying decision of a customer is also influenced by organisational factors.
Every global organisation is recognised by it’s objectives and goals, an
efficient organizational structure and a well acknowledged system and producer
for buying. Directly and indirectly these factors influence a person’s buying decision.
These distinctive features provide us with reasons for determining the buying
decision.

4.     
Interpersonal: The buying decisions in the
industries are usually cumulative and also in accordance to the
rules/regulations decided. The buying centres needs varied individuals with
definite status, formal authority and forcefulness. The buying centres are made
of the individuals of an organization bothered with purchase decision method.
The risk originating out of it is shared among them.A common goal is shared
among them. There is communication among the members of a buying centre
regarding the purchases to be made. During the marketing buying decision there
is a chance of clash among the members of a buying centre. The
marketing/purchasing programme is adjusted accordingly after the suppliers know
about such conflicts and resolve them justifiably .The disputes between the
buying centre participants has to be solved quickly so that the activity of
purchasing shall be done promptly i.e. as per the prepared schedule of
production and sales.The marketer should have a deep knowledge of ‘group
dynamics’ to help them while settling conflicts and setting early release of
purchase order.

 

2.2
Clustering Algorithms

Clustering
can be understood by as a concept based on “learning by observation”. It is an
unsupervised learning method and does not require a training data set to
generate a model.Clustering helps us in the discovery of previously unknown
groups/clusters within the data.The grouping is done in such a way that the
objects within the same group (cluster) are very like/similar to each other but
they are very unlike/dissimilar to the objects in some other group. Machine
Learning algorithms improves the buying strategy by facilitating them with
more extensive and broad data. Random forest algorithm comes into the
category of supervised classification algorithm. This algorithm creates a
forest consisting of a number of trees, as the number of trees in the
forest increases more potent and powerful the forest becomes. Similarly,
in the random forest classifier, the higher the number of trees in
the forest the higher the accuracy results are 4.
Table 1 shows the comparison between different techniques.

 

 

Table1: Comparison of
different Techniques3

 

 

 

 

Technique

Feasibility for huge data set

Dependence on iteration

High speed

Input Specified by user(No. of clusters)

Applicability for all data based

Agglomerative (Hierarchical)

û

ü   

û

ü   

û

Density based

û

ü   

ü   

û

û

EM

ü   

ü   

ü   

ü   

û

K- Mean

ü   

ü   

ü   

ü   

û

 Random Forest

ü   

û

û

û

ü   

 

 

 

3. Methodology

 

In
this paper random forest algorithm is used to predict behavior of customers and
formulate four marketing plans and strategies in order to develop long term
customer relationships and compare their behavior from actual data and then
calculate accuracy. Figure1 shows the architecture of classification customer
behavior. There are seven features of data set where the random forest
algorithm is applied. After applying the random forest, different trees have
been generated which has been classified into targeted value.

 

 

 

Figure 1: Architecture of
classifying customer behavior

 

Figure 1 shows the architecture of
classifying the customer behavior where the dataset with different features set
is classifying using random forest algorithm.

 

Algorithm:

For D= 1 to n where n represent the
number of respondent
 for B = 1 to m  where m represent the features
Trees are the total no. of trees produced
for  i=1 to B
                                           Choose
sample Di form D
                                           Construct
tree Ti using D:
At each node , choose subset of
features
and only consider splitting on these
features
save the output of each tree in Trees
Take majority voted value(Trees)
Return voted value
 
 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 2: Algorithm for
classification

Figure 2 shows the
random forest algorithm for classification. Table 2 shows the behavior assigned
corresponding to the shopping pattern.

 

Table 2: shows the
behavior assigned corresponding to shopping pattern

 

Behaviour Assigned

Shopping
Pattern

 

Home office

Corporation

Small
Business

Consumer

Furniture

Office supply

Technology

 Interpersonal

ü   

û

û

û

ü   

û

û

Individual

û

û

ü   

ü   

û

ü   

û

Environmental

û

ü   

û

û

û

û

ü   

Organizational

û

ü   

û

û

û

ü   

û

 

 

 

 

4.         Implementation and Result

 

The classification of   consumer
purchasing is implemented in the language python.  Data set has pre-defined categories or
segments for customer classification. Data is collected from Kaddle
data.com.  Data is divided into different
categories such as customer segment, product sub category.  The customer segment is again divided in to
different sub categories such as Home office, Corporate, small Business and consumer. These
product categories are divided into
categories such as furniture, office supplies, and technologies. The product
subcategories are again divided into book cases, tables, chairs etc.

 

The survey was made to know about the preferences and choices
of customers in different regions. Questions were asked
to respondents about importance
of price while choosing any product, priority of service;
location etc. Respondents have to fill the choice accordingly. Survey was conducted
by online and offline mode. This survey goes through the general mass.
On the basis of this survey the category of behavior is assigned to customer.

Figure 3 shows the reasons
for shopping. It has been seen that 72% customer is interested to buy furniture
for corporate where as 22 percent customer buys furniture for personal use.
Only one percent customer buys furniture for small business.

Figure 3:    Shopping Pattern

 

Figure 4 shows the number of user buy
different type of products. 

Figure 4 : Bar graph depicting
behavior of the people

 

 

 Figure 5 represent the different bar graph for
different region.  In these graphs, a
comparison is shown between the predicted and actual behaviors of particular
regions. Behavior is depicted on the X-axis whereas Y-axis represent the number
of respondent. Bar graph with blue color shows predicted behavior i.e. output
from random forest algorithm whereas green color represent the actual behavior
which is taken from data set. These four regions are central, west , east and
mountain.

 

          

Figure 5 : Comparision between actual and prdicted behavior for
different region

Table 3 shows the actual
and predicted behavior of all the four behavior pattern. Table 4 shows the
precision and recall for different behavior.

Actual                                            Predicted

Environmental

Individual

Organizational

Interpersonal

405

17

0

0

41

201

2

0

1

3

203

0

0

0

0

288

            Table
3: Actual and Prediucted behavior
of different pattern

Table
4: Precision and Recall for
different behavior

Categories

Precision

Recall

F1-
score

Support

Environmental

0.91

0.96

0.93

422

Individual

0.91

0.82

0.86

244

Organizational

0.99

0.98

0.99

207

Interpersonal

1.00

1.00

1.00

288

 

Result shows that after applying the random
forest algorithm the accuracy is 0.94487107666.

 

 

Conclusion

 The result
shows that the random forest algorithm has achieved 94% accuracy to predict the
consumer behavior. This algorithm gives very good
accuracy as compared to other existing algorithms.
Different type and larger data sets can use to predict
the behavior of consumer in
upcoming years.

References

1 Manali Khaniwale, “Consumer Buying
Behavior,” International Journal of Innovation and Scientific Research,
vol. 14, no. 2, pp. 278–286, April 2015.

2 Kumar, A. H., John, S. F., &
Senith, S. , “A Study on factors influencing consumer buying behavior in
cosmetic Products”,  International Journal of Scientific and Research
Publications, 4(9), 1-6, 2015.

3 Abbas, O. A., “Comparisons Between Data
Clustering Algorithms” International Arab Journal of Information
Technology (IAJIT), 5(3), 2008.

4 Ghatasheh, N., ” Business analytics
using random forest trees for credit risk prediction: A comparison study”,
 International Journal of Advanced Science and Technology, 72, 19-30,
2014.