Machine Learning based Consumer Trend Analysis Using

Fast Associativity Analysis of Products

1.1 Introduction
The study of the person who wants to buy certain product from shopping place, at the time of
purchasing product, why he wants to buy it? This is very interesting concept to study. To
studyhis psychological mindset and converting this into statistical format and seethat is there any
technical format by which we can analyze his buying behavior.
The study of consumers helps firms and organizations to improve their marketing strategies by
understanding issues such as how
1) The psychology of consumers that how he thinks,feel reasons and select between
different alternatives.
2) The mindset of how the consumer is influenced by his or her environment.
3) The behavior of consumer while shopping or making other marketing decisions.
4) How customer motivation and decision strategy differ between products that differ in
their level of importance or interest that they entail for the customer; and
5) How management can adjust and improve their marketing campaigns and marketing
ideas to more effectively reach customer.
1.2 Consumer trendAnalysis
Consumer trend means the study of individuals, groups or organizations about their process of
selecting, securing, using and disposing the products, services, experiences or ideas to satisfy
needs and the impact of this process on the consumer and the society.
Behavior concerns either with the individual or thegroup (e.g. In college friends influence what
kind of clothes a person should wants to wears) or a firm (peoples working in firm make decision
asto which products the firm should use.) The use of product is often so important to the
marketer because this may influence how a product is best positionedor how we can encourage
increased consumption.Consumer trend involves services and ideas as well as tangible products.

The application of Consumer trend is widely acknowledged in making of marketing strategy i.e.
for making bettermarketing campaigns. For example by understanding that children’s are more
receptive to food advertising i.e. when they are watching the cartoonfilms after their school
timing, they are also hungry, so we learn to schedule snack advertisements late in the afternoon.
A second application is public policy. There are many deaths occurred due to chewing of tobacco
products as well as smoking. So the government i.e. Federal drug administration (FDA) made it
compulsory to tobacco manufacturing companies to print the warning message on the tobacco
packets, pouch and containers. Social marketing involves getting ideas across to consumers
rather than selling something.
As a final benefit, studying Consumer trend alsocorrelate consumers common sense, e.g. if you
buy a one litter bottle of ketchup for 65 Rs, at the same time you pay higher, when you purchase
half litter bottle. However if you often pay a size premium by buying the larger quantity. In other
words in this case knowing these facts will sensitize you to the need to check the unit cost labels
to determine if you are really getting a bargain

Fig 1: General Working of Consumer Trend Analysis

Trend Prediction is also known as association rule mining or affinity analysis, is a data-mining
technique that originated in the field of marketing and more recently has been used effectively in
other fields, such as bioinformatics, nuclear science, pharma co epidemiology, immunology, and
geophysics. The goal of Trend Prediction is to identify relationships (i.e., association rules)
between groups of products, items, or categories. We describe Trend Prediction and explain that
it allows for inductive theorizing; can address contingency (i.e., moderated) relationships; does
not rely on assumptions such as linearity, normality, and residual equal variance, which are often
violated when using general linear model based techniques; allows for the use of data often
considered “unusable” and “messy” in management research (e.g., data not collected specifically
for research purposes); can help build dynamic theories (i.e., theories that consider the role of
time explicitly); is suited to examine relationships across levels of analysis; and is practitioner
friendly. We explain how the adoption of Trend Prediction is likely to help bridge the much
lamented micro macro and science practice divides. We also illustrate that use of Trend
Prediction can lead to insights in substantive management domains, such as human resource
management (e.g., employee benefits), organizational behavior (e.g., dysfunctional employee
behavior), entrepreneurship (e.g., entrepreneurs’ identities), and strategic management (e.g.,
corporate social responsibility). We hope our article will serve as a catalyst for the adoption of
Trend Prediction as a novel methodological approach in management research.

Trend Prediction allows researchers to uncover non obvious and usually hidden and
counterintuitive associations between products, items, or categories. This methodological
approach allows researchers to identify those items that co-occur (i.e., appear together) on a
frequent basis and assess the extent to which they co-occur. Trend Prediction has been used to
understand consumer behavior regarding types of books that are purchased together (as
purchased on Amazon.com) as well as different types of wines that the same individual is likely
to purchase (as purchased on VirginWines.com). Because Trend Prediction originated in the field
of marketing and was initially used to understand which supermarket items are purchased
together (i.e., placed together in the same “basket”), the technique adopted the name market
basket analysis.
Seem to have used Trend Prediction first. Agrawal et al. are computer scientists who had access
to a large repository of previously collected customer transaction data and were able to discover
association rules between items purchased. The method was quickly adopted in the field of
marketing as a popular tool for a variety of practical applications. For example, suppose that a
marketing researcher uses Trend Prediction to gather empirical support that cake mix and icing
tend to be purchased together. The researcher may then determine that cake mix and icing are
complementary items (i.e., categories) such that lowering the price of only one of the goods is
associated with an increase in demand for both goods. From a theoretical standpoint, Trend
Prediction can be used to develop hypotheses and theories inductively. For example, these results
suggest the hypothesis that consumers may have mental models that include associations among
many other sets of items that are complementary in terms of their activities and interests (e.g.,
running shoes and water bottles). From a practical standpoint, this study’s results can be used to
make decisions such as stocking the two items near each other, thereby increasing the likelihood
that customers will easily find and purchase both products instead of just one of them.
The applicability of Trend Prediction is not limited to analyzing archival data. The method can
also be used with primary data. For example, mailed out surveys in which respondents completed
open-choice checklists regarding common food allergens. As a result, Kanagawa et al. found that
certain food allergens tend to occur together in the same person. As a second illustration of the
use of Trend Prediction with primary data, collected responses from secondary school students in
Singapore about their preferences regarding the use of school-based counseling services. On the
basis of their responses to the administered questionnaires, students were separated into two
groups—those willing to see a counselor and those unwilling and goh and Ang identified the
main traits and concerns of each group. In short, although Trend Prediction has traditionally been
used with archival data, it can also be used with primary data.
Categorical association rules are based on binary data, and this is the most common type of
association rules because Trend Prediction was originally developed to analyze shopping cart
data in binary form: Either a customer bought or did not buy a particular product. However, it is
possible to derive association rules involving continuous variables. Association rules involving a
continuous variable are called quantitative association rules. Categorical association rules relate a
value of a categorical variable with a value of another categorical variable, but a quantitative
association rule relates a value of a categorical variable with a summary statistic of a continuous
variable (e.g., mean, median). As an example, used Trend Prediction to derive quantitative
association rules based on hotel customers. The data set included responses to Likert-type scales
assessing behavioral and attitudinal constructs provided by customers of a large hotel chain.
Bennedich derived several quantitative association rules, such as those relating certain floor
numbers with mean room satisfaction scores. In short, Trend Prediction allows for the derivation
of association rules including not only categorical but also continuous variables, such as those
measured using multiple-item scales.
1.3 Advantages of Using Trend Prediction in Management Research Trend
Prediction
1.3.1 Allows for Inductive Theorizing
Trend Prediction is a powerful technique for inductive theorizing. More specifically, Trend
Prediction allows researchers to make use of vast stores of data for theory-building purposes.
The importance of the inductive capability of Trend Prediction was highlighted by Locke (2007),
who argued that inductive approaches are essential for the development of theory. The impact
and influence of inductive theorizing are also supported empirically by the finding that theory-
building elements in an empirical article lead to a greater number of citations compared to the
presence of theory-testing elements. Using Trend Prediction has the potential to lead to important
contributions by allowing researchers to implement an inductive approach to theory building,
which, in spite of its advantages, is currently underutilized in management research.
Indeed, using Trend Prediction for inductive theory building has led to insights in marketing and
other fields. For example, pointed out that the use of Trend Prediction has allowed marketing
researchers to build theoretical models of purchasing decisions involving products in more than
one category (i.e., multiple-category decision making). Other researchers have also used Trend
Prediction inductively to study a number of important questions outside the field of marketing.

Trend Prediction can also be used inductively to generate hypotheses that are then tested
deductively in follow-up work. For example, collecting archival data from a car rental company
to derive association rules that related certain types of phrases uttered by customer service agents
to whether customers later picked up reserved cars. To test whether certain phrases did indeed
improve the probability of pickup, conducted a follow-up experiment to draw causal inferences.
Customer service agents were randomly assigned to two groups: those who were trained on the
recommendations based on the derived association rules (treatment) and those who were not
(control). The dependent variable was average pickup ratio, defined as the ratio of the number of
actual pickups to the number of reservations. Results were analyzed using a general linear model
(GLM) based technique (i.e., two independent-group ttests) and indicated that the average
improvement in the pickup ratio of the treatment group was greater than that of the control
group. In short, this study offers a good illustration of the possibility of first using Trend
Prediction to inductively generate hypotheses and then conducting a follow-up study that is
deductive in nature and uses GLM-based data-analytic procedures.
1.3.2 Trend Prediction Can Address Contingency Relationships
Trend Prediction can be used to uncover contingency relationships, also labeled moderating or
interaction effects. Specifically, Trend Prediction can reveal the strength of a given relationship
as well as the extent to which such relationships (i.e., association rules) vary across different
contexts. For example, suppose that using Trend Prediction yields the association rule linking
variables A and B from a data set. Once a binary moderator (consisting of groups g1 and g2) is
added to the data set, Trend Prediction may reveal that the confidence of [(A), (B)], or the
association rule relating A and B, is weaker for g1 and stronger for g2. In addition, Trend
Prediction may reveal that a distinct association, [(C), (B)], is also an association rule for g1 but
not for g2 and that yet another distinct association, [(F), (A)], constitutes an association rule for
g2 but not g1.

1.3.3 Trend Prediction Does Not Rely on Often Untenable Assumptions

Trend Prediction is not bound by the strict, and often untenable, assumptions such as the
linearity, normality, and residual equal variance required by the GLM and other frequently used
data analytic methods in management research. Trend Prediction assesses relationships between
items or categories as opposed to linear relationships between two or more variables. As a result,
Trend Prediction is free from the strict assumptions that are often violated in management
research.
Clearly, as is the case for any methodological approach, Trend Prediction does have certain
requirements that must be met. We describe these requirements later in the sections titled “Steps
Involved in Using Trend Prediction” and “Potential Pitfalls in the Use of Trend Prediction.”

1.3.4 Trend Prediction Allows the Use of “Unusable” and “Messy” Data
Trend Prediction enables researchers to use data that are often seen as “unusable” for
management research. Most organizations regularly collect data on many business functions,
such as human resources. For example, given the availability and affordability of data storage
systems, organizations regularly collect data on employees (e.g., performance, absenteeism,
benefits choices, training opportunites), customers (e.g., purchasing choices, frequency of visits),
and many other issues. Moreover, such data are often collected unsystematically, sporadically,
and without a particular scientific study in mind. Trend Prediction is ideally suited to be used
inductively with such data sets to uncover association rules that may not be readily apparent.
“Messy” data often involve issues such as missing values and outliers. This type of data is
common and poses many challenges in management research, particularly at the macro level of
analysis. While Trend Prediction is not immune to the problem of missing values, it allows for
the interpretation of missing data as indicating that no option was selected or preferred. As
previously discussed, Trend Prediction does this by deriving association rules with values of lift
below 1.0, such that these association rules use the presence of one item to predict the absence of
another item. Alternatively, an Trend Prediction open-source software called Weka
(http://www.cs.waikato.ac .nz/~ml/weka/) contains a function called Decision Stump that treats
missing values as distinct items that Trend Prediction can either predict or use to predict other
items, thereby deriving information from both the presence and the absence of data. Thus, Trend
Prediction allows researchers to derive association rules such as the following: If a customer
orders a tofu vegan meal as the main dish, there is a 65% chance that the customer will not order
any dessert. A caveat to this advantage is that missing responses, or non choices, must be
substantively meaningful, as opposed to caused by artifacts, for association rules that treat the
absence of an item as a distinct item to be meaningful (see the section titled “Step 4: Check
Trend Prediction Requirements”).
Another advantage of Trend Prediction related to messy data is that association rules are less
influenced by outliers compared to more traditional data-analytic approaches. In the context of
Trend Prediction, outliers result in infrequently occurring associations. For example, results may
indicate that only one customer’s “basket” included a washing machine and a warranty for a
computer. Such a combination, whether the result of an error in data entry or a customer
behavior that is extremely uncommon, will have a lift value near unity and very small values for
support and confidence. Consequently, this outlier will not influence overall substantive results
to the extent that an outlier would influence GLM results based on standard errors, correlation
coefficients, and regression coefficients.
We fully acknowledge that no method can overcome errors in data collection or entry.
Nonetheless, Trend Prediction offers researchers some degree of flexibility while attempting to
make full use of an extant data set. Researchers bound by GLM’s assumptions will find many
data sets unusable; despite the value those data sets might offer as a window into organizational
phenomena that may be otherwise unavailable.

1.3.5 Trend Prediction Can Help Build Dynamic Theories

The field of management is increasingly acknowledging the important role of time in theory
building (Mitchell & James, 2001). Accordingly, there is a growing interest in understanding
phenomena as they unfold longitudinally. Another advantage of Trend Prediction is that it can
help build theories from not only cross-sectional but also longitudinal data. Thus far, we have
made no distinction regarding when the transactions occur. However, Trend Prediction allows
for theory-building efforts that are both cross-sectional and longitudinal (i.e., dynamic) in nature.
Given the nature of the available data, there are mainly two ways of building dynamic theories
via Trend Prediction:
(a) multiple Trend Prediction and
(b) sequential Trend Prediction (Trend Prediction). The multiple Trend Prediction approach is
used when the available data include transactions as they have occurred over time (i.e.,
transactions at Time 1, transactions at Time 2, . . . , transactions at Time k). The multiple Trend
Prediction approach involves first treating each data wave as a single cross-sectional study and
then examining whether the lift, support, and confidence of association rules vary over time,
which can be done largely descriptively by using graphs. Trend Prediction is used when the
available data describe individual events (i.e., items) as they have occurred over time. For
example, a data set may consist of employees’ files since the moment they were hired. So, the
data set may include information gathered during their first, second, third, and so forth, month of
employment. Trend Prediction may uncover the presence of an association rule such as
suggesting that there is a pattern in which Event A (e.g., being assigned to a formal mentor)
occurs before B (e.g., volunteering to participate in a particular training and development
program), which occurs before C (e.g., receiving higher performance evaluation scores).

1.3.6 Trend Prediction Can Be Used to Assess Multilevel Relationships

Trend Prediction can be applied across all levels of analysis ranging from within-individual to
firm-, industry-, and country-level contexts. For example, Trend Prediction can be used to
simultaneously examine categories representing individual as well as group, team, or
organizational characteristics. In the field of marketing, if the organizational unit is used as the
level of analysis, then researchers can derive association rules linking lower-level (e.g.,
individual Subway restaurants) with higher-level units (e.g., entire chain of Subway restaurants
within a particular state) and even higher levels of analysis (e.g., all Subway restaurants within
the United States).
Methods for assessing relationships between lower-level predictors and higher-level outcomes
are nascent and in early stages of development. However, many management theories posit such
effects. Using Trend Prediction has great potential to lead to theoretical advancements regarding
the presence of such relationships.

1.3.7 Trend Prediction Is Practitioner Friendly

The widely documented science–practice divide in the field of management is due, at least in
part, to a lack of good communication between researchers and practitioners, who seem to speak
different languages. Trend Prediction is practitioner friendly because of how results are
presented. Numerical results produced by Trend Prediction are intuitive and easy to understand
from a practical significance perspective. For example, support and confidence are based on a
probability scale from 0% to 100%, which facilitates interpretation of association rules’ practical
significance. In other words, Trend Prediction relies on the strength of the association rules,
which makes results more easily understandable by practitioners.
Another indication of its practitioner friendliness is that, over the course of its development,
Trend Prediction has proved to be particularly useful because it offers actionable insights. Recall
how immediately applicable recommendations were derived from our earlier example involving
purchase transactions of soda, orange juice, fat-free cream, and donuts. When researchers are
able to use Trend Prediction in their own work and collaborate with practitioners to produce
insights that are relevant and actionable in terms of applications, both sides benefit, and as a
result, we may see a narrowing of the science practice divide.
In short, Trend Prediction has great potential as a methodological approach in the field of
management because it allows for inductive theorizing, can address contingency relationships,
does not rely on assumptions that are often violated when using GLM-based techniques, allows
for the use of data often considered “unusable” and “messy,” can help build dynamic theories, is
suited to examine relationships across levels of analysis, and is practitioner friendly. Next, we
provide a step-by-step description of how to use Trend Prediction.

1.4 Literature Survey

Lang, T., & Rettenmeier, M. (2017). , Understanding consumer behavior with recurrent
neural networks Consumer behavior in e-commerce can be described by sequences of
interactions with a webshop. We show that recurrent neural networks (RNNs) are a natural fit for
modeling and predicting consumer behavior. In multiple aspects, RNNs over advantages over
existing methods that are relevant for real-world production systems. Applying RNNs directly to
sequences of consumer actions yields the same or higher prediction accuracy than vector-based
methods like logistic regression. Unlike the latter, the application of RNNs comes without the
need for extensive feature engineering. In addition, we show that RNNs help us link individual
actions directly to predictions in an intuitive way. This allows us to understand the implications
consumer actions have on predicted probabilities over the course of the consumer's history. We
demonstrate the advantages of RNNs on the empirical data of a large European online fashion
platform.

Santolaya, D. S. (2017). Using recurrent neural networks to predict customer behavior

from interaction data, Customer behavior can be represented as sequential data describing the
interactions of the customer with a company or a system through the time. Examples of these
interactions are items that the customer purchases or views. Recurrent Neural Networks are able
to model effectively sequential data and learn directly from low-level features without the need
of feature engineering. In this work, we apply RNN to model this interaction data and predict
which items the user will consume in the future. Besides exploring how effective are RNNs in
this scenario, we study how item embeddings can help when there is a high quantity of different
items, providing a comparison and analysis of different methods to learn embeddings. Finally,
we apply attention mechanisms to gain interpretability in the RNN model. We compare different
variants of attention mechanism, providing their performance and the usefulness to explain the
predictions of the model.

Bhargav, A. et al, in "Market basket analysis using artificial neural network" 2014 [4], the
authors describe Market basket analysis is based upon the identification and analysis of
purchasing patterns of the customers. The problem with market basket analysis is the varying
needs of the customers with respect to seasons and time and so they need to perform it time and
again. Another problem that arises while doing market basket analysis is with Apriori algorithm
in which they need to find candidate sets and frequent item-sets time and again. In this paper,
they are suggesting the use of artificial neural network technique to overcome these problems.
They have used single layer feed-forward partially connected neural network technique for this
purpose.

Abbas, W.F. et al, in "Discovering Purchasing Pattern of Sport Items Using Market Basket
Analysis" 2013 [5], the authors describe One of the oldest problem in data mining is the market
basket problem, the search for meaningful associations in customer purchase data. Currently, the
Sport Company has an issue on sport items arrangement in accordance with customer purchasing
pattern. They noticed that, the sales of certain products become decrease when they made some
arrangement to the shelves. The Sport Company do not have any available computerized
mechanism to provide the best arrangement of item store at the retail store. Everything is done
manually by the owner of the shop according their own style. This study intends to identify
purchasing pattern of sport items by adopting data mining technique which is Market Basket
Analysis. This data mining pattern will help the retailer to make a better arrangement of the
products at the premise. Historical data is analyzed to identify associated items from purchasing
data of customer that involved sales data, items data and order data. As a result from this
research, the sports items will be arranged according to the best rules identified and propose a
new pattern.

Nengsih, W. in "A comparative study on market basket analysis and apriori association
technique" 2015 [6], the authors describe Association Rules is one of the data mining
techniques which is used for identifying the relation between one item to another. Creating the
rule to generate the new knowledge is a must to determine the frequency of the appearance of the
data on the item set so that it is easier to recognize the value of the percentage from each of the
datum by using certain algorithms, for example apriori. This research discussed the comparison
between market basket analysis by using apriori algorithm and market basket analysis without
using algorithm in creating rule to generate the new knowledge. The indicator of comparison
included concept, the process of creating the rule, and the achieved rule. The comparison
revealed that both methods have the same concept, the different process of creating the rule, but
the rule itself remains the same.

Trnka, A. in "Market Basket Analysis with Data Mining methods" 2010 [7], the authors
describe This paper describes the way of Market Basket Analysis implementation to Six Sigma
methodology. Data Mining methods provide a lot of opportunities in the market sector. Basket
Market Analysis is one of them. Six Sigma methodology uses several statistical methods. With
implementation of Market Basket Analysis (as a part of Data Mining) to Six Sigma (to one of its
phase), they can improve the results and change the Sigma performance level of the process. In
their research they used GRI (General Rule Induction) algorithm to produce association rules
between products in the market basket. These associations show a variety between the products.
To show the dependence between the products they used a Web plot. The last algorithm in
analysis was C5.0. This algorithm was used to build rule-based profiles.

Schluter, T. et al, in "About the analysis of time series with temporal association rule
mining" 2011 [8], the authors describe This paper addresses the issue of analyzing time series
with temporal association rule mining techniques. Since originally association rule mining was
developed for the analysis of transactional data, as it occurs for instance in market basket
analysis, algorithms and time series have to be adapted in order to apply these techniques
gainfully to the analysis of time series in general. Continuous time series of different origins can
be discretized in order to mine several temporal association rules, what reveals interesting
coherences in one and between pairs of time series. Depending on the domain, the knowledge
about these coherences can be used for several purposes, e.g. for the prediction of future values
of time series. They present a short review on different standard and temporal association rule
mining approaches and on approaches that apply association rule mining to time series analysis.
In addition to that, they explain in detail how some of the most interesting kinds of temporal
association rules can be mined from continuous time series and present an prototype
implementation. They demonstrate and evaluate their implementation on two large datasets
containing river level measurement and stock data.
Schwenke, C. et al, in "Analysis and simulation of sales receipt data in supermarkets" 2011
[9], the authors describe For the prediction of product sales, simulations of interactions of
different individual customer behaviors in supermarkets are an appropriate approach. In order to
parameterize the simulation close to reality, analysis of shopping baskets based on real sales
receipt data is one of the key steps. The combination of the data analysis and simulation is in the
scope of this paper Analysis of real sales receipt data was done by determining so-called
customer prototypes, which are representatives of different customer classes. In order to make
the simulation insensitive against short transient changes, longitudinal analysis and clustering
was carried out, additionally to the common cross section analysis. Longitudinal analysis was
necessary to calculate similarities between shopping basket clusters of different points in time.
The resulting stabile customer prototypes then were used to parameterize a supermarket
simulation. As result, this paper shows the analysis of consumer data as well as the
parameterizing and validation of a supermarket simulation. The combining of data mining with
simulation results in a better accuracy of analysis and the ability to evaluate influencing factors
that can not be extracted directly through data mining.

Gupta, N. et al, in "An implementation and analysis of DSR using market basket analysis
to improve the sales of business" 2014 [10], the authors describe The mention system is
implementing to enhance the sales of retail business. Daily Status Report system is use for
increasing the performance of retail business by using the data mining algorithms. In this system
they use the market basket analysis to find out the frequent patterns. In this system they are able
to track the location of employee. This system also enables to access the daily basis information
of the product sales with particular areas. Therefore company administrator can easily know the
performance of particular employee. And also he is able to know the sale of particular product if
the sale of particular product is low then for increasing the sale of the product company include
offer in that product. In the mention system they use the market basket analysis technique which
shows the purchasing habits of the customer.

Laura Maria Badea, “Predicting Consumer Behavior with Artificial Neural Networks”,
2014 [11]Nowadays, facile access to information and advancements in processing power unfold
opportunities for new decision support techniques used for financial and economic purposes.
Artificial neural networks are machine learning techniques which integrate a series of features
upholding their use in financial and economic applications. Backed up by flexibility in dealing
with various types of data and high accuracy in making predictions, these techniques bring
substantial benefits to business activities. This paper investigates how consumer behavior can be
identified using artificial neural networks, based on information obtained from traditional
surveys. Results highlight that neural networks have a good discriminatory power, generally
providing better results compared with traditional discriminant analysis.

1.5 Problem Formulation

In the Consumer trend mining area, most of the research efforts went in the first place to
improving the algorithmic performance and in the second place into reducing the output set by
allowing the possibility to express constraints on the desired results. Over the past decade a
variety of algorithms that address these issues through the refinement of search strategies,
pruning techniques and data structures have been developed. While most algorithms focus on the
explicit discovery of all rules that satisfy minimal support and confidence constraints for a given
dataset, increasing consideration is being given to specialized algorithms that attempt to improve
processing time or facilitate user interpretation by reducing the result set size and by
incorporating domain knowledge.

There are also other specific problems related to the application of association rule mining from
e-learning data. When trying to solve these problems, one should consider the purpose of the
association models and the data they come from. Nowadays, normally, data mining tools are
designed more for power and flexibility than for simplicity. Most of the current data mining tools
are too complex for normal users to use and their features go well beyond the scope of what
anuser might require.

1.6 Proposed work

Over the years Data mining (DM) can used to understand the consumer buying behavior using
various techniques. Data mining has gradually increases many folds and today it is a giant 100-
billion-dollar industry. In data mining world every activity of a consumer in a supermarket is
treated as a byte of data. How the consumer spends, which day what time normally he/she does
the shopping, what they buy most often,how much they buy, in that locality etc. All this data
which is gathered somewhere at the backend about which a consumer is not even aware and
there is a big industry which is slicing & dicing this data & selling it at a premium price.
Most of the time the data is used to analyze the pattern or shopping habits of consumers like in
festive season which product sells more. What are the associations between these products? Is
there anypattern in this habit? If data show some common theme then stores management
arranges that product accordingly. e.g. If management arrange electronic product like television,
LCDs, Tape recorders, Mobiles etc. withattractive schemes in the 9 front row in festive season.
And also arrange the similar items which customer tends to buy along with these product.

To make moreprofit stores will not run any discount or special offers on the products on busy
days. Yet another area commonly traced is the weekly shopping habit of thecustomer, what
products they buy and of what quality. This information can be used for stocking purposes and
handle the inventory cost. Likewise there are many other aspects in which this data analysis is
leading to better consumer satisfaction. For monthly analysis about the certain product demand
i.e. buying in the start of the month and buying at the end of the month? The people have money
to spend in the starting of the month and at the end of the month people spend less. In the
vacation of the school and at the starting of the school the requirement of certain commodity is
increase. So to maintain the inventory and also to increase sell in this period. It is absolutely
necessary to grab this opportunity of consumer needs and implement novel algorithms that
can help us mine the big data produced by consumers and trends.

1.7 Objectives
1. To study and Analyze Various Consumer trend analysis methods and algorithms
2. To collect Market basket data (big data) forpredictingconsumer trends.
3. Implement new fast and more accurate consumer trend prediction algorithm.
4. To study and analyses algorithms performance using metrics such as time, complexity
and items mined etc.
1.8 Tools and Techniques
For the above said research work, we will be requiring to access the various e-journals like IEEE,
Science Direct and ACM.For data collection and analysis of the data, I will be requiring various
software’s and Internet facility. The hardware and software requirements for my thesis work are
Hardware Requirements:
 Processor: Pentium IV or Higher
 RAM: I GB or Higher
 Hard Disk: 10 GB or Higher
Software Requirements:
 Platform: Windows7 Professional
 Java 8, Netbeans 7 and above
 Microsoft office( Excel and Word)

1.9 Methodology
Association rule mining finds interesting association or correlation relationships among a large
set of data items, the researcher is become more interested in mining association rules from
organizations database. The discovery of interesting association relationship among huge
amounts of customer transaction records can help in many business decision making processes
such as catalog design, cross marketing and loss-leader analysis.
We will implement Market Basket Analysis for big data analysis because Market Basket analysis
is a tool of knowledge discovery about co-occurrence of nominal or categorical items. Market
Basket Transaction or market Basket Analysis is a data mining technique to derive association
between data sets. Researcher has categorical data of transaction records as input to the analysis
and the output of the analysis is association rules as a new knowledge directly from data.
Researchers have following transactional data from an organization and the numbers of
transactions in one day are limited as the data shown below.
 Transaction Items from the customer who bought
ID more than one item
1 Sugar, Wheat, pulses, Rice
2 Sugar, Pulses
3 Wheat, Pulses
4 Pulses, Wheat, Rice
5 Wheat, Pulses
6 Sugar, Wheat
7 Sugar, Rice, Pulses

Based on the data above, we can derive the following output of association rule by using market
Basket analysis.

Output- Association Rules

People who bought this Support Confidence

item Also bought the following items
Wheat Pulses 57% 80%

Rice Pulses 43% 100%

The association rule will have the following form

X→Y
that form has meaning that people who bought items of set x are often also bought items on set Y
e.g. if X = { sugar, wheat} and Y ={Rice, Pulses} and we get association rule indicates that
people who bought sugar and wheat also bought Rice and pulses.
Work Plan

Jan, 2018 March, 2018 April, 2018 May, 2018 June, 2018
Time Activity to to to to to
Feb, 2018 April, 2018 May, 2018 June, 2018 July, 2018

Literature review
&
Synopsis

To study and
propose WSN
Energy Model

To design WSN
Scenarios

To Simulate and
Compare WSN
Energy Models

Article(s)/paper(s)
Writing

Thesis writing
 References
[1]. Lang, T., & Rettenmeier, M. (2017). Understanding consumer behavior with recurrent
neural networks. In International Workshop on Machine Learning Methods for
Recommender Systems.

[2]. Santolaya, D. S. (2017). Using recurrent neural networks to predict customer behavior
from interaction data.
[3]. Wu, Xindong, Xingquan Zhu, Gong-Qing Wu, and Wei Ding. "Data mining with big
data." Knowledge and Data Engineering, IEEE Transactions on 26, no. 1 (2014): 97-107.
[4]. Stephen Kaisler, Frank Armour, J. Alberto Espinosa, and William Money. "Big data:
Issues and challenges moving forward." In System Sciences (HICSS), 2013 46th Hawaii
International Conference on, pp. 995-1004. IEEE, 2013.
[5]. R. Saranya, V. P. Muthukumar, and N. Mary. "BIG DATA IN CLOUD COMPUTING."
[6]. Bhargav, A., Mathur, R.P., Bhargav, M., "Market basket analysis using artificial neural
network", Sch. of Comput. Eng., Lovely Prof. Univ., Phagwara, India,
10.1109/I2CT.2014.7092091, 1-6, 2014

[7]. Abbas, W.F., Ahmad, N.D., BintiZaini, N., "Discovering Purchasing Pattern of Sport
Items Using Market Basket Analysis", Fac. of Comput. & Math. Sci., Univ. Teknol.
MARA, Shah Alam, Malaysia, 10.1109/ACSAT.2013.31, 120-125, 2013

[8]. Nengsih, W., "A comparative study on market basket analysis and apriori association
technique", Dept. of Comput., Politek. Caltex Riau, Riau, Indonesia,
10.1109/ICoICT.2015.7231468, 461-464, 2015

[9]. Trnka, A., "Market Basket Analysis with Data Mining methods", Dept. of Appl. Inf.,
Univ. of SS. Cyril & Methodius, Trnava, Slovakia, 10.1109/ICNIT.2010.5508476, 446-
450, 2010

[10]. Schluter, T., Conrad, S., "About the analysis of time series with temporal association rule
mining", Inst. of Comput. Sci., Heinrich Heine Univ., Dusseldorf, Germany,
10.1109/CIDM.2011.5949303, 325-332, 2011
[11]. Schwenke, C., Vasyutynskyy, V., Kabitzsch, K., "Analysis and simulation of sales
receipt data in supermarkets", Dept. of Tech. Inf. Syst., Dresden Univ. of Technol.,
Dresden, Germany, 10.1109/ETFA.2011.6059082, 1-8, 2011

[12]. Gupta, N., Yadav, M.L., "An implementation and analysis of DSR using market basket
analysis to improve the sales of business", Dept. of Comput. Sci. & Eng., Amity Univ.,
Noida, India, 10.1109/CONFLUENCE.2014.6949249, 82-86, 2014.

[13]. Badea, Laura Maria. "Predicting consumer behavior with artificial neural networks."
Procedia Economics and Finance 15 (2014): 238-246.