Journal of Physics: Conference Series

PAPER • OPEN ACCESS

Recent citations
Online Store Product Recommendation System - HPAI consumer shopping analysis using
Uses Apriori Method Apriori algorithm
Y Findawati et al

To cite this article: C S Fatoni et al 2018 J. Phys.: Conf. Ser. 1140 012034

View the article online for updates and enhancements.

This content was downloaded from IP address 60.243.48.24 on 29/07/2021 at 13:03

IC-ELINVO IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 1140 (2018) 012034 doi:10.1088/1742-6596/1140/1/012034

Online Store Product Recommendation System Uses Apriori

Method

C S Fatoni1, E Utami2 and F W Wibowo3

1,2,3
Magister of Informatics Engineering, AMIKOM Yogyakarta University, Indonesia.

E-mail: fatoni.work@gmail.com, emma@nrar.net, ferry.w@amikom.ac.id

Abstract. Online Store grows very fast. Online Store helps people to buy the desired product
online. Heavy competition among of Online Store provider give a rise to technology
development. Many Online Store system not only display the product but also need to be
supported by proper products selection to attract the attention of website visitors. As a result,
many website visitors are confused when they are going to buy products in Online Store. The
number of product variety offered to a customer when he buys goods sometimes more than one
product. The problem leads to an idea of developing a products recommendation system.
Historical data from visitors and customers can be used to analyze the user needs and products
preferences. Association rule using Apriori knowledge will be able to capture the user
preference. By identify the user preferences, a valid product recommendation can be developed.
This research will analyze the rules in a historical data of purchase from Online Store visitors
to get recommendation of products to be displayed. According to experiment result, the
ascociation rule is capable to generate precise recommendations with confidence values
76.92%.

1. Introduction

The Internet becomes a very important medium for its users to share information and conduct business
transactions by the company. Improving the service and business of the company becomes the main
factor to develop online media as a means of conducting online transactions. The development of the
field of online media business one of them is Online Store, it allows users to buy goods online. "Forget
B2B (Business to Business) or B2C (Business to Customer), business is about P2P (Path to
Profitability) - the path to profitability or profit" [1]. Information from around the World; eMarketer;
about the Online Store share of total global retail sales from 2015 to 2021 is shown in Figure 1 [2].

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
IC-ELINVO IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 1140 (2018) 012034 doi:10.1088/1742-6596/1140/1/012034

Figure 1. Total E-Commerce Global Retail Sales Graph.

The growing Online Store market growth is a major concern for business people to develop
technology to make it easier for customers to buy the products they want. The growth of the digital
economy the greater the business competition will also get heavier, consequently more and more
Online Store not only display product products but also need to be supported by the selection of the
right products to attract the attention of website visitors. The number of product variations offered to
customers when purchasing goods is sometimes more than one product, consequently many customers
are confused choosing the desired product, even there are customers who buy various products and per
transaction.

The problem of confused customers when buying a product leads to the idea of developing a product
recommendation system. Historical data from visitors and customers can be used to analyze user needs
and product preferences. Ascociation rules using Apriori knowledge will capture user preferences. By
identifying user preferences, valid product recommendations can be developed. This research will
analyze the rules in the historical data of Online Store visitor's purchase to get product
recommendation to be displayed. Apriori method implements taking data with associative rules
(Association rule) to determine the associative relationship of item combination [3]. Association rules
are implemented through mechanisms of counting the support and trust of item relations. Association
rules are interesting if the value of support is greater than the minimum support and also the value of
the confidence is greater than the minimum trust. This a priori algorithm would be appropriate to
apply when there are several link items to be analyzed.

Historical data about purchases from visitors and Online Store customers can be used to form item
combination data, in testing data whether item combinations meet the minimum support and
confidence parameters. When item combinations meet parameters, they can be used to determine valid
product recommendations for customers who want to buy products.

2. Related Works

Section on the topic can be more than one according to the needs of the author. The content of the
topic section according to the author's needs. For the section "topic", the author gives the title
according to the needs of the author and the number of topics according to the needs of the author. If
there is only one topic, then the section number follows to be: 1. Introduction, 2.Topic (title according
to author's need), 3. Conclusion, 4. References.

2
IC-ELINVO IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 1140 (2018) 012034 doi:10.1088/1742-6596/1140/1/012034

The last remaining research (Ali Anaissi, and Madhu Goyal, 2015) raises the title of the SVM-Based
Association Rules for Knowledge and Classification. The ultimate goal is to find an existing
relationship between product sales that can help retailers identify new opportunities to sell their
products to customers. This paper aims to find hidden patterns of knowledge in large data sets that can
generate more understanding for data holders and identify new opportunities for important tasks
including strategic planning and decision making. Researchers try to implement data mining tools that
will allow analysts to interact with data and answer business questions such as promotional
advertising. This research produces a list of association rules which are then grouped into different
categories based on product results from the rules. SVM is used to test whether this group can capture
variations based on different types of rules. Significant classification accuracy is achieved by SVM,
which allows clients to target customers with their needs and propose a useful campaign [4].

Subsequent research (Ezhilarasan C, and Ramani S., 2017) research entitled Predicted performance
using Modified Clustering Techniques with Fuzzy Association Rule Mining Approach to Retail
classifies data based on components and related equations, using rules of mining fuzzy rules for
implementation. Grouping is done on one object and set in one cluster, when overlapping and has
multiple clusters for a new object using the fuzzy method. The size of the proximity is determined by
the matrix. Analysis of this system is made to improve sales predictions better, and performance based
on association rules using fuzzy model. In this study data were extracted from Online Store to predict
sales by type of group [5].

Several studies related to information systems have been carried out by (Yusuke Ito, and Shohei Kato,
2016) research entitled An Apriori-Based Approach to Product Placement in Order Picking make and
manage warehouse goods as easy as possible, the purpose is to shorten the time in the collection of
goods in the warehouse, in this study written that the method applied is very successful and very
effective, but this research is only done on small-scale warehouses, not on a large scale [6].

3. Methodology

The research has several stages ranging from literature studies to experimental methods and through
the process of testing data mining. The system design method used is Reuse-Based has 6 stages in the
system design process, among others, the collection of system specification requirements, component
requirement analysis, system specification modification, combining the system design with Reuse-
Based, development of merger system, and system validation process. The concept approach allows
for the retrieval of reusable components and depends on the size of components that can be reused and
integrated with the concept of components in the software, described in Figure 2 [7].

Figure 2. Reuse Based Designing Method.

3.1. System Requirement Collection

The process of collecting system requirements in this study covers the need for Online Store related
information, data needs from problem identification, problem solving, database retrieval, data
processing planning taken from database in realtime so that it will be processed into data mining.

3
IC-ELINVO IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 1140 (2018) 012034 doi:10.1088/1742-6596/1140/1/012034

3.2. Component Requirement Analysis

The analysis will be done after gaining access to the Online Store database then performing data
retrieval periodically, with date filter, month month. Grapheap data is a key component in data mining,
if the data has been obtained then the steps to determine the database to store all the data, and then
create a data matrix for the calculation using the a priori method.

3.3. Modification of System Specifications

Modification of system specifications that can be done in data mining, after making the process of data
retrieval will be processed in a different database with existing Online Store system, it will create its
own system integrated with Online Store.

3.4. Component Requirement Analysis

The application of Reused-Based is more to initialization of system requirements at the beginning,
then the data retrieval is always updated, as well as the implementation of the process of retrieving
historical data purchase transactions on Online Store on the system for data mining. Data mining is a
process that uses statistical, mathematical, artificial intelligence, and machine learning techniques to
extract and identify useful information and gather knowledge from large databases or data warehouses
[8]. Based on the definition that has been submitted, the important things related to data mining are
data mining is an automated process for existing data, data to be processed in the form of very large
data, and the purpose of data mining is to obtain relationships or patterns that may provide useful
indications.

The next process is the analysis using the priori method aims to find the combination of items with the
frequency pattern of the results of transactions on Online Store. To process the a priori method
described in figure 3.

Figure 3. Reuse Based Designing Method.

In Figure 3 the priori method will continue if the data condition is still possible for the next itemset. If
not possible then soon be the result. The customer purchase transaction matrix will take the number of
transactions from each product item per transaction and the amount of transaction data then used to
determine the combination of product items.

Apriori method is one algorithm that searches for frequent itemset by using rule association technique.
The Apriori method uses knowledge of the frequency of previously known attributes for processing
more information. In the Apriori algorithm determines possible candidates with respect to minimum
support and minimum trust. Support is the visitor value or the percentage of item combinations in the
database.

The main processes undertaken in the Apriori method for obtaining frequent itemsets include: (1) Join
(merge): This process is done by combining items with other items until no more combinations can be

4
IC-ELINVO IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 1140 (2018) 012034 doi:10.1088/1742-6596/1140/1/012034

established; and (2) Prune (pruning): The trimming process is the result of a merged item then
trimmed by using the minimum support specified by the user. The principles of the Apriori Algorithm
include: (1) Collect single items then look for the biggest item; (2) Get a candidate pair and then count
the major pair of each item; (3) Find the candidate triplets of each item and so on; and (4) Every part
of the frequent itemset should be frequent.

3.5. System Development and Integration

The system development process and its integration involve retrieving data for processing in data
mining and then applying a priori method to determine the combination of item item items, which will
then be processed by a system integrated with Online Store and displayed in a data format as per
system requirements analysis for recommendations customer products.

3.6. System Validation Process

System validation refers to the system test of the data processed in this study compared to the data
processed by using the system created, for validation of the desired results will be tested gradually by
using more data to obtain recommendations from more results.

4. Result and Discussion

4.1. System Design Analysis
The process of system design analysis in accordance with the steps in the flow of Online Store product
recommendation system, following the flow of system design analysis process described there is
Figure 4.

Figure 4. Design of Process Flow Analysis System

The process of system design analysis begins with data collection related to customer purchase
transactions in Online Store. The data obtained are then analyzed and presented in the historical data
of Online Store visitor purchases, with associative rules for determining the associative relationship of
a combination of product items purchased by the customer, the calculation method using the
mechanism of counting the support and beliefs of a relationship item. Association rules are interesting
if the value of support is greater than the minimum support and also the value of the confidence is
greater than the minimum trust. Apriori algorithm is applied when there are several items of
relationships that want to be analyzed. Testing data mining processing then from the results of product
output recommendations appropriate for customers will be done gradually by looking at the next
customer purchase history data in a certain period.

4.2. Data collection

The initial process of system design analysis is to collect customer purchase history data taken at the
end of the year from June to July 2017.

Table 1. Pattern of Customer Purchase Transaction

Trx Item Transaction

1 Shoes, Jacket, Pants, Shirts, Hijab
2 Hands, Bags, T-shirts, Shirts, Pants
3 Clothes, Shoes, Bags, Wallets, Pants
4 Perfumes, Shirts, Bags, Glasses, Shoes
5 Wallet, Bag, Jacket, Jas, Perfume

5
IC-ELINVO IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 1140 (2018) 012034 doi:10.1088/1742-6596/1140/1/012034

Trx Item Transaction

6 Belts, Watches, Eyeglasses, Handbags
7 Bags, Watches, Shirts, Dress, Hijab
8 Jas, Shirts, Pants, T-shirt, Shoes
9 Pants, Wallet, Glasses, Hijab, Shoes
10 T-shirts, Belts, Pants, Shirts, Shoes
11 Pants, Jackets, Bags, Sweater, Hijab, Shirts
12 Sweaters, T-shirts, Watches, Dresses, Pants, Hijab
13 Hijab, Shirts, Pants, Sweaters, Shoes, Wallets
14 Hijab, Perfume, Shoes, Shirts, T-shirts
15 Shoes, Wallets, Belts, Eyeglasses
16 Jackets, Shoes, Wallets, Bags, Pants
17 Clothes, Shoes, Watches, Shirts, Pants
18 Hijab, Bags, Gowns, Perfumes, Shoes
19 Watches, Shoes, T-shirts, Pants, Hijab, Bags
20 Wallet, Perfume, Pants, Eyeglasses, T-Shirt

4.3. Transaction Data Matrix

Customer purchase transaction data will be adjusted in matrix format to get the pattern of goods
formed, and used to know how many times the goods are purchased by the customer on each
transaction, the following data is presented in table 2.

Table 2. Customer Purchase Transaction Matrix

Trx Item transactions

Wristwatch

Eyeglasses
Perfume

Sweater
T-shirts
Jacket
Shoes

Dress
Purse

Hijab
Pants

Shirt

Belt
Bag

Jas

1 1 1 1 0 0 0 1 0 0 0 0 0 0 1 0
2 0 0 1 1 1 1 1 0 0 0 0 0 0 0 0
3 1 0 1 0 1 0 1 0 0 1 0 0 0 0 0
4 1 0 0 0 1 0 1 1 1 0 0 0 0 0 0
5 0 1 0 0 1 0 0 1 0 1 0 1 0 0 0
6 0 0 0 1 1 0 0 0 1 0 1 0 0 0 0
7 0 0 0 1 1 0 1 0 0 0 0 0 0 1 1
8 1 0 1 0 0 1 1 0 0 0 0 1 0 0 0
9 1 0 1 0 0 0 0 0 1 1 0 0 0 1 0
10 1 0 1 0 0 1 1 0 0 0 1 0 0 0 0
11 0 1 1 0 1 0 1 0 0 0 0 0 1 1 0
12 0 0 1 1 0 1 0 0 0 0 0 0 1 1 1
13 1 0 1 0 0 0 1 0 0 1 0 0 1 1 0
14 1 0 0 0 0 1 1 1 0 0 0 0 0 1 0
15 1 0 0 0 0 0 0 0 1 1 1 0 0 0 0
16 1 0 1 0 1 0 0 0 0 1 0 1 0 0 0
17 1 0 1 1 0 0 1 0 0 0 0 0 0 0 1
18 1 0 0 0 1 0 0 1 0 0 0 0 0 1 1
19 1 0 1 1 1 1 0 0 0 0 0 0 0 1 0
20 0 0 1 0 0 0 1 1 1 1 0 0 0 0 0

6
IC-ELINVO IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 1140 (2018) 012034 doi:10.1088/1742-6596/1140/1/012034

4.4. Analysis of Itemset Creation

The customer purchase transaction matrix will take the number of transactions from each product item
per transaction and the amount of transaction data then used to determine the itemset combination.

4.4.1 Itemet Combination Analysis. The combination of 1 itemset is processing based on the data provided in
table 2, the process of forming K1 or called a combination of 1 itemset with the minimum amount of support =
40%, by the formula in equation (1) [9].

The squatter process in equation (1) obtains the data shown in table 3, for the support value of each
product item.
Table 3. The support value of each product item
Item Quantity Support
Shoes 13 65%
Jacket 3 15%
Pants 13 65%
Wristwatch 6 30%
Bag 10 50%
T-shirts 6 30%
Shirt 12 60%
Perfume 5 25%
Eyeglasses 5 25%
Purse 7 35%
Belt 3 15%
Jas 3 15%
Sweater 3 15%
Hijab 9 45%
Dress 4 20%

The establishment of itemsets in table 3 with a minimum of 40% support can be found that meets the
minimum standards of support on shoe product items, pants, watches, bags, shirts, wallets. Then from
the result of combination formation 1 item will be done combination 2 itemset as in table 4.

4.4.2. Itemet Combination Analysis. The Combination of 2 Items is processing based on data provided in
table 3 items taken above the support value of each product item, the process of forming K2 or called a
combination of 2 itemsets with minimum amount of support = 40%, by the formula in equation (2) [9].

The squatter process in equation (1) obtains the data shown in table 3, for the support value of each
product item.
Table 4. Support value from a combination of 2 product itemsets
Item Quantity Support
Shoes, Pants 9 45%
Shoes, Bags 5 25%
Shoes, Shirts 8 40%
Shoes, Hijab 6 30%
Pants, Bags 5 25%

7
IC-ELINVO IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 1140 (2018) 012034 doi:10.1088/1742-6596/1140/1/012034

Item Quantity Support

Pants, Shirts 10 50%
Pants, Hijab 6 30%
Bags, Shirts 4 20%
Bags, Hijab 4 20%
Clothes, Hijab 4 20%

Combination 2 itemsset with minimum 40% support can be seen combination of 2 itemset that meet
minimum standard of support that is shoes, pant with support of 45% and pants, shirt with 50%
support. From the result of the combination of 2 itemsset will be done formation 3 itemsset as in table
5.
4.4.3 Itemet Combination Analysis. Combination of 3 Items is processed based on data provided in table 4
items taken above the support value of each product item, the formation process K3 or called with a combination
of 3 itemsets with minimum amount of support = 40%, by the formula in equation (3) [10].

Table 5. Support value from a combination of 3 product items

Item Quantity Support
Shoes, Pants, Shirts 6 30%

4.5. Association Rules

After all the high frequency patterns are found, then the association rules that meet the minimum
requirements for confidence by calculating the trust of the associative rule P → Q [10]. Minimum
Confidence = 70%. The confidence value of rule P → Q is obtained.

Table 6. Support value from a combination of 3 product items

Rules Quantity Confidence
If buying shoes, buy pants 9/13 69.23%
If buying a pair of pants, buy a T-shirt 10/13 76.92%

Based on Table 6, the products most often purchased by customers are shoes, Pants, T-shirts with
knowledge of the products most often purchased by customers, then the company can develop
strategies in determining the purchase of products to maintain product availability required by
customers and also can adjust the the location of the product based on the combination of product
items formed.
4.6. System Implementation
The authors of this study took sample sample data from June to July 2017 to provide product
recommendations in December 2018. Based on this transaction the authors took the highest trust value
as the basis for determining the right product recommendations for Online Store customers. Figure 5
shows the product brand and the type of product being sold.

8
IC-ELINVO IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 1140 (2018) 012034 doi:10.1088/1742-6596/1140/1/012034

Figure 5. Product Form and Product Data Brand

Figure 6. Product Transaction Form by Product Type

Figure 6 shows the transaction data of the type of product purchased from June to July 2017. From this
product type data by applying the a priori method by applying a minimum support value of 40%, and a
minimum value of 70% confidence generating confidence values from various itemset combinations.
Based on the value of itemet combination with the highest confidence value, the system can provide
product recommendations to be promoted or sold to Online Store customers in Figure 7.

Figure 7. Output Form Generate Product Recommendations

9
IC-ELINVO IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 1140 (2018) 012034 doi:10.1088/1742-6596/1140/1/012034

4.7. System Testing

Calculation with priori method in table 6 and from calculation with built system is determined
minimum value of support: 40% and minimum trust value 70% then generated tendency of product
purchased by customer is presented in table 7.

Table 7. The calculation results

Calculation Technique Results
Calculation using Apriori method If buying a pair of pants then buy a
t-shirt with a trust value of 76.92%
Calculations using the built system If buying a pair of pants then buy a
t-shirt with a trust value of 76.92%

Testing using a priori method and using the built system yields the same value in accordance with the
minimum value of support and minimum trust that has been determined. From these results then this
system can help the pattern of product display recommendations based on the tendency of Online
Store products purchased by customers consisting of 1 item of the product, then from the test results
can also help management Online Store in providing product recommendations to customers and as
desired.

5. Conclusion

Based on the discussion of research that the authors do, it can be concluded that by applying the a
priori algorithm, the system provides product recommendations to Online Store customers based on
the trust value of a combination of products purchased at a given time period. Application of Apriori
Method in this research is to find the most combination of items based on transaction data and then
form the association pattern of item combination. This research can be further developed using a
weighted product method to get the weight of each product brand so that it can provide an alternative
to the most recommended product brands.

Acknowledgments
Universitas AMIKOM Yogyakarta for the support by giving the facility which is needed for the study.
This research also can be done by a good cooperation among all sector.

6. References

[1] Berman D K 2000 Dot Coms: Can They Climb Back? (Business Week online June 26th)
[2] eMarketer 2018 Information from around the World: 2015 to 2017
(https://www.statista.com/statistics/534123/e-commerce-share-of-retail-sales-worldwide)
[3] Aprizal, Hasriani, and Wahyu N 2015 Implementasi Data Mining Untuk Menentukan Posisi
Barang Pada Rak Menggunakan Metode Apriori Pada PT. Midi Utama Indonesia (Techno.
COM) Vol 15 p 335-342.
[4] Anaissi A and Goyal M 2015 SVM-Based Association Rules for Knowledge Discovery and
Classification, 2015 2nd Asia-Pacific World Congress on Computer Science and
Engineering (APWC on CSE)
[5] Ezhilarasan C and Ramani S 2017 Performance Prediction using Modified Clustering
Techniques with Fuzzy Association Rule Mining Approach for Retail 2017 International
Conference on Intelligent Computing and Control

10
IC-ELINVO IOP Publishing
IOP Conf. Series: Journal of Physics: Conf. Series 1140 (2018) 012034 doi:10.1088/1742-6596/1140/1/012034

[6] Ito Y and Kato S 2016 An Apriori-Based Approach to Product Placement in Order Picking
International Conference on Agents
[7] Fatoni C S and Noviandha F D 2018 Case Based Reasoning Diagnosis Penyakit Difteri dengan
Algoritma K-Nearest Neighbor Citec Journal Vol 4 No 3 p 220-232
[8] Turban E 2005 Decision Suport Systems and Intelligent Systems (Yogyakarta: Andi Offset)
[9] Yanto R and Khoiriah R 2016 Implementasi Data Mining dengan Metode Algoritma Apriori
dalam Menentukan Pola Pembelian Obat Citec Journal Vol 2 No 2 p 102-113
[10] Ummi K 2016 Analisa Data Mining Dalam Penjualan Sparepart Mobil Dengan Menggunakan
Metode Algoritma Apriori (Medan: CSRID) Vol 8 No 3 p 145-208

11