[PDF] [PDF] Product Inventory Predictions at Small Medium Enterprise - CORE

[PDF] The Marriage of Market Basket Analysis to Predictive Modeling

Market basket analysis (MBA); predictive modeling; curse of dimensionality; rules -based (RB) dimensional reduction schemes; RB input discovery MBA REVIEW

[PDF] Product Inventory Predictions at Small Medium Enterprise - CORE

Study conducted by Joseph, Pratikto, Gerry (2006) prove that Market Basket Analysis (MBA) is a reliable method in understanding the customer buying behavior 2 1 Prediction of Inventory MBA is a method in data mining that focus on identification of products that are purchased at the same time on each transaction

[PDF] Market Basket Analysis - CORE

Market Basket Analysis(MBA) also known as association rule learning or affinity Prediction: In this functionality, prediction of some unknown or missing 

Product Inventory Predictions at Small Medium - ScienceDirect

Study conducted by Joseph, Pratikto, Gerry (2006) prove that Market Basket Analysis (MBA) is a reliable method in understanding the customer buying behavior 2 1 Prediction of Inventory MBA is a method in data mining that focus on identification of products that are purchased at the same time on each transaction

Market Basket Analysis - ScienceDirectcom

Market Basket Analysis(MBA) also known as association rule learning or affinity Prediction: In this functionality, prediction of some unknown or missing 

[PDF] Market Basket Analysis - RUN - Universidade NOVA de Lisboa

Data mining; Market basket analysis; association rules; apriori algorithm; changing market basket analysis is being used to build predictive models and to get 

[PDF] Market Basket Analysis in Retail - UPCommons

A complete process of generating a predicting model could be the same as the one used in a cooking recipe Ingredients would be the data and the algorithm the 

[PDF] Using Market Basket Analysis in Management - Herman Aguinis

Market basket analysis (MBA), also known as association rule mining or affinity analysis, is a Forecasting cyclones is a difficult task because of their tendency 

[PDF] Personalized Market Basket Prediction with - ISTI Open Portal

Index Terms—Next Basket Prediction, Temporal Recurring Sequences, User- Centric Model, Market Basket Analysis, Data Mining, Interpretable Model ♢

[PDF] preface for physics project class 12

[PDF] prefblog

[PDF] preferential certificate of origin philippines

[PDF] preferential rate hotel meaning

[PDF] prefill customs form

[PDF] prefletter

[PDF] prefrontal cortex decision making

[PDF] prefrontal cortex function

[PDF] pregnancy book in telugu pdf

[PDF] pregnancy book malayalam pdf

[PDF] pregnancy day by day book pdf free download

[PDF] pregnancy guide book in bengali pdf

[PDF] pregnancy guide for first time mothers pdf

[PDF] pregnancy guide for new moms

[PDF] pregnancy planner pdf

Procedia Economics and Finance 4 ( 2012 ) 312 - 320

2212-5671 © 2012 The Authors. Published by Elsevier Ltd.

Selection and peer-review under responsibility of Parahyangan Catholic U niversity. doi: 10.1016/S2212-5671(12)00346-2 International Conference on Small and Medium Enterprises Development with a Theme (ICSMED 2012) Product Inventory Predictions at Small Medium Enterprise Using

Market Basket Analysis Approach - Neural Networks

Agus Mansur

a* , Triyoso Kuncoro a a

Industrial Engineering Department ,Universitas Islam Indonesia, Jl. Kaliurang km 14, 5 Yogyakarta 55584, Indonesia

Abstract

One of the key problems in every company, including small and medium enterprises, is how to determine the inventory

level for each product that will be sold to their customers appropriately as it can suppress the build up of inventory as well

as avoid the stock out. This study is aimed to understand the behavior of consumers in purchasing the products so it can be

used to predict the purchasing for the next period. Later, the prediction is used as a decision support in determining the

appropriate amount of inventory for each product. The study was conducted at Karomah Brass, a small and medium

enterbuys from the supplier. The methods that used in this study are the Market Basket Analysis (MBA) and Artificial Neural

Network (ANN) Back propagation. MBA is used to examine the buying behavior of customer while ANN Back customers frequently purchase products that serve as a kind of antique closet accessories and if customer bought that certain

product, then they will also buy similar products in accordance with 21 rules that have been obtained from the mining of

transaction data. Whereas, other result shows the prediction of the amount product inventory requirements/needs for one

year to the next.

© 2012 The Authors. Published by Elsevier Ltd.

Selection and/or peer-review under responsibility of Parahyangan Catholic University. Keywords: customer behavior; inventory level; prediction; MBA; ANN; Backpropagation.

Corresponding author.

E-mail address: gus_mansur@yahoo.com

Available online at

www.sciencedirect.com

© 2012 The Authors. Published by Elsevier Ltd.

Selection and peer-review under responsibility of Parahyangan Catholic U niversity.Open access under CC BY-NC-ND license.

Open access under CC BY-NC-ND license.brought to you by COREView metadata, citation and similar papers at core.ac.ukprovided by Elsevier - Publisher Connector

313 Agus Mansur and Triyoso Kuncoro / Procedia Economics and Finance 4 ( 2012 ) 312 - 320

1. Introduction

Inventories are defined as raw materials, works in process (WIP) or the finished products that are stored to

meet the demand (Herjanto, 1999; Baroto, 2002). If the amount of inventory is less than the amount of actual

need, the company will loose the opportunity to maximize their sales, gain new customers, get customer

loyalties and gain the maximum profits. While, if they stock too much inventory, it will increase the costs of

maintenance and storage, so that it will reduce the profits (Sari, 2010). Because of the great impact of inventory

their biggest investments (Winata & Abbas, 2008).

Problems that ment

other products are stock out. These problems occur beca use KB can not read the patterns of their customer

buying behavior. This study aims to answer these problems, especially in understanding the customer buying

behavior and prediction of product inventory needs for the next period.

2. Customer Buying Behaviour

David & Bitta (1998) defined customer buying behavior as a decision process that makes people choose and

use the products or services. Customer buying behavior is not easy to find out because of the human behavior

itself. Study conducted by Joseph, Pratikto, & Gerry (2006) prove that Market Basket Analysis (MBA) is a

reliable method in understanding the customer buying behavior.

2.1 . Prediction of Inventory.

Prediction, especially in the field of inventory, has attracted the attention of researchers and practitioners

nowadays. Many studies have been conducted, one of them is Pujihastuti (2008). Pujihastuti uses moving

average method in a company that has a fluctuating demand and prove that MA is able to accommodate rapid

changes in information and suitable with the condition of the company that has a high variety of products and

raw materials. However, this method is less appropriate when used to predict long-term predictions (Arsyad,

2001).

Another studies on the application of another forecasting methods are such as Exponential Smoothing (Tanuwijaya, 2008), Brown method (Winata & Abbas, 2008), Exponential Smoothing Winter (Paramita &

Tanuwijaya, 2010), and Box-Jenkins (ARIMA) (Naibaho, 2009), but due to the limitations of each method

above, it is difficult to apply in the KB. Pratama (1999) uses ANN Backpropagation to predict inventory and

proves that ANN offers so many advantages, it is very accurate and the computation can be done with computer in an efficient algorithm so that the calculation does not take much time.

Based on the previous studies above, the ANN method is seen as the most appropriate method to predict the

amount of product inventory needs in KB, so the prediction inventory process in this study will only focus by

using this method.

2.2 Market Basket Analysis

MBA is a method in data mining that focus on identification of products that are purchased at the same time

on each transaction. Output of MBA is a set of rules that indicate the products that are purchased on the same

time. This output will be used as input for the prediction of inventory. More detail, the rules generated by MBA are associatio n rules that have form "If antecedent (A), then

consequent (B)". each rule is equipped with a support level that indicates the number of transactions containing

314 Agus Mansur and Triyoso Kuncoro / Procedia Economics and Finance 4 ( 2012 ) 312 - 320

A and B and confidence level that is a measure of accuracy which is the rule of association rules. Each rule is

also equipped with an expected condition and a lift. For each antecedent (A) and consequence (B), the support,

confidence, expected confidence, and lift are formulated as follows (Yusuf, Pratikto, & Gerry, 2006):

Support = Number of transaction contains A & B/Number of transactions [1] Confidence = Number of transaction contains A & B/Number of transactions contain A [2] Expected Confidence = Number of transaction contains B/Number of transactions [3]

Lift = Confidence/Expected confidence [4]

2.3 Artificial Neural Network (ANN)

ANN is a model that mimics the workings of biological neural networks and has been widely applied to the

prediction problems (Eliyani, 2007). Backpropagation (BP) is an ANN that has capability to train a network to

be able to give correct response based on the input pattern (Maru'ao, 2010). It is necessary to select input data

(P) or target (T) for this study because the training process that is used is supervised training. The learning

algorithms used in ANN training for this study is the Levenberg Marquardt algorithm. We choose this algorithm because this algorithm is faster (Kusumadewi, 2004).

3. Research Methodology

The research methodology of this study briefly describe in figure 1 below. Start

Problems Identification

End

Literature Study

Data Colection

Data Processing:

1. MBA

2. ANN Backpropagation

Discussion and Conclution

A A

Fig. 1. Research Methodology

We use software named XL miner to handle the MBA process, as shown in Fugure 2.

315 Agus Mansur and Triyoso Kuncoro / Procedia Economics and Finance 4 ( 2012 ) 312 - 320

No.

Trans Product Freq

items sets (support) Product Probability (confidence) Product

1 H13 Malang,

Black couplers

K, H13 3 H13 Malang, Black

couplers K, H13 100% H13 Malang,

Black couplers

K, H13

2

Keyhole, Key 1.5

cm 5

Keyhole, Key 1.5 cm

100%

Keyhole, Key

1.5 cm

3

Keyhole, Key 1.5

cm 3

Keyhole, Key 1.5

cm 100%

Key 1.5 cm

Fig. 2. MBA process

There are two steps on the Association Rule Algorithm (Tama, 2010), first find the Frequent Item sets

(probability of combination that frequently appear from a set of items). Second, find the association rules of a set of Frequent Item sets. Start

Compute the

feedforward End

Compute the Jacobian

matrics

Compute the changes in

the weights and biases

Recompute the

feedforward

Recompute the MSE

New MSE

Min? No Yes

Fig. 3. Detailed of ANN Backpropagation

316 Agus Mansur and Triyoso Kuncoro / Procedia Economics and Finance 4 ( 2012 ) 312 - 320

Fig. 4. Network Architecture

In the fig. 4., shown that the network only has 1 input cell unit, because it only uses 1 variable, namely

number of product. The network also has 1 hidden layer with 5 cell units, it is intended to speed up the

computation process. The output layer only has 1 output cell, because it only has 1output variable, namely

number of product.

The Matlab commands are shown as below:

Input and Target Data

P = [160 259 28 146 90 36 108 184 189 110 161 94]; T = [165 261 30 150 95 40 110 190 195 115 165 100];

Building the Feed Forward Network

net = newff([28 259],[5 1],{'tansig' 'purelin'},'trainlm');

Training Process

net = train(net,pn,tn) TRAINLM-calcjx, Epoch 0/1000, MSE 0.311041/0.001, Gradient 0.363677/1e-010 TRAINLM-calcjx, Epoch 1/1000, MSE 0.300253/0.001, Gradient 0.838133/1e-010 TRAINLM-calcjx, Epoch 2/1000, MSE 0.071804/0.001, Gradient 0.226616/1e-010 TRAINLM-calcjx, Epoch 3/1000, MSE 0.0692973/0.001, Gradient 0.334378/1e-010 TRAINLM-calcjx, Epoch 4/1000, MSE 0.0151635/0.001, Gradient 0.472178/1e-010 TRAINLM-calcjx, Epoch 5/1000, MSE 0.00220436/0.001, Gradient 0.0363479/1e-010 TRAINLM-calcjx, Epoch 6/1000, MSE 0.00109866/0.001, Gradient 0.0118473/1e-010 TRAINLM-calcjx, Epoch 7/1000, MSE 0.000788252/0.001, Gradient 0.00680927/1e-010

TRAINLM, Performance goal met.

317 Agus Mansur and Triyoso Kuncoro / Procedia Economics and Finance 4 ( 2012 ) 312 - 320

Fig. 5. Network Training Process

In the Fig. 5. shown that the network training process is stop at the 7 epoch, it because the MSE of network

training (valued 0.000788252) < error goal that has been defined (valued 0.001). It means that the network

performance is well. In the fig. 6. shown that the Target graphic is closed to the Output graphic, it means that

a = Columns 1 through 10

166.7375 255.9069 35.6269 151.5184 91.1725 41.7922 110.0364 191.8998 196.9273 112.1806

Columns 11 through 12

167.8133 95.2874

Fig. 6. Comparison of Target and Output

318 Agus Mansur and Triyoso Kuncoro / Procedia Economics and Finance 4 ( 2012 ) 312 - 320

4. Discussion and Conclusion

The data that we used for this research were 564 sales transaction data on a small enterprise that sell antique

furniture accessories, named Karomah Brass, starting from the September 2010 through August 2011. Table 1

contains examples of the data.

Table 1 Example of the raw data

Date Trans

No Receipt

No

Qty Description

Unit Price

(Rp) Total Price (Rp) September 02, 2010 1 124 8 Iron Hinges 3" 3500 28000

4 Q3 K 7500 30000

4 K8 5000 20000

2 125 6 K4 4500 27000

6 K2 3000 18000

3 126 12 Iron Hinges 2" 8000 96000

Date Trans

No Receipt

No

Qty Description

Unit Price

(Rp) Total Price (Rp) September 10, 2010 4 127 100 Thick Hinges 3 " 19000 1900000

96 Thick Iron Hinges 3 " 7000 672000

5 128 6 Key 2.5 cm 7700 46200

6 H13 7500 45000

10 Mini Bolt 1800 18000

Later, the sales transactions data were processed using MBA technique. We used a software named XL Miner Version 3 to generate the rules. We set 3 as the minimum support value and 80% as the minimum confidence. The results were 21 rules, as shown in the table 2 (example) below. From the rule, we can draw the conclusions about the customer buying behavior at KB, such as the

also buy keyhole and key 1.5 cm. From these knowledge, we can suggest to the KB to offer black couplers and

H13 to any customer that buy H13 Malang, and so on.

Table 2. The example of MBA generated rules

Rule # Conf.

Antecedent (a) Consequent (c)

Support

(a) Support (c) Support (a U c)

Lift Ratio

1 100

H13 Malang, Black couplers

K

H13 3 26 3 21.69231

2 100 Key 1.5 cm 5 60 5 9.4

3 100 Key 1.5 cm 3 60 3 9.4

4 85.71 Iron bolt K. Keyhole Key 1.5 cm 7 60 6 8.057143

5 100 H13. H13 Malang Black couplers K 3 83 3 6.795181

From our observations, all of the products in the generated rules were antique cabinet accessories so that

they were the most frequent products purchased by consumers. We took these items as our main concern for

the next step, inventory prediction, because the most frequent product must be never in stock out condition if

319 Agus Mansur and Triyoso Kuncoro / Procedia Economics and Finance 4 ( 2012 ) 312 - 320

bolt B, iron bolt K, mini bolt, H10, H13, H13 Malang, K1, K2, K4, klik klok, key 1.5 cm, key 2.5 cm, keyhole,

P8 K, black couplers K, and white couplers.

Later, we used ANN backpropagation to predict the need of these items for next year, starting from

September 2011 to August 2012. We used ANN backpropagation for the predictions. The network architectures

were 1 neuron of input layer , 1 layer with 5 neurons of hidden layer, and 1 neuron of output layer. The MSE

ue to the zero because if we set it to zero, the maximum epoch will be reached before we reach the MSE goal, the result is shown in table 3 below. Table 3. The quantity predictions of each item based on ANN Backpropagation

No Items Qty

1 D8 30

2 Iron bolt B 49

3 Iron bolt K 48

4 Mini bolt 54

5 H10 33

6 H13 54

7 H13 Malang 15

8 K1 117

9 K2 302

10 K4 282

11 65

12 Key 1.5 cm 162

13 Key 2.5 cm 232

14 Keyhole 295

15 P8 K 37

16 Black couplers K 1622

17 White couplers 437

5. Future Works

This study is limited to these limitations:

We are not considering the cost of each the product procurement from the supplier. We are not considering the costs that occur as long as the product is in inventory. We are not considering the selection of neural network architectures to be the optimal.

So that, we suggest to next research to consider the procurement cost and apply the costs that occur because of

inventory is employed as the input to the determination of the optimum amount of product that will be bought and stored.

References

Arsyad, L., (2001). Peramalan Bisnis (Business Forecasting). Yogyakarta: BPFE-Yogyakarta.

Baroto, T., (2002). Perencanaan dan Pengendalian Produksi (Production planning and controlling) . in T. Pudjadi, & Iwan, 2009. Aplikasi

Sistem Informasi Persediaan pada PT. Panca Pipando (PPI) dengan Metode Distribution Resource Planning (Application of

inventory information system at PT Panca Pipando using distribution resource planning). Seminar Nasional Aplikasi Teknologi

Informasi (SNATI).

David, L. L., & Bitta, A. J., (1998). Consumer Behaviour. in L. Deswindi, 2007. Kecepatan Tingkat Penerimaan dan Perilaku Konsumen

Terhadap Produk Lama yang Mengalami Perubahan dan Produk Inovasi Baru dalam Upaya Memasuki dan Merebut Pasar

(Acceptance rates and customers behaviour at the innovation of the maturity products compared with the new ones). Business &

320 Agus Mansur and Triyoso Kuncoro / Procedia Economics and Finance 4 ( 2012 ) 312 - 320

Management Journal Bunda Mulia

, 3, 20.

Eliyani. 2007. Peramalan Harga Saham Perusahaan Menggunakan Artificial Neural Network Dan Akaike Information Criterion (Stock

price prediction using artificial neural network and aikake information criterion). Seminar Nasional Aplikasi Teknologi (SNATI).

Herjanto, E. (1999). Manajemen Produksi dan Operasi (Operation management and production) . in W. Winata, & B. S. Abbas, 2008.

Sistem Informasi Persediaan Bahan Baku dengan Continuous Review System (Information system of inventory using continuos

review system). Jurnal Piranti Warta, 11.

Kusumadewi, S. (2004). Membangun Jaringan Syaraf Tiruan Menggunakan Matlab & Excel Link (Developing artificial neural network

using Matlab and Excel Link)). Yogyakarta: Penerbit Graha Ilmu.

Maru'ao, D. O. (2010). Neural Network Implementation in Foreign Exchange Kurs Prediction. Gunadarma University, Faculty of Industrial

Engineering.

Naibaho, P. (2009). Kajian Perencanaan Produksi Agregat di PT. Wiska (Agregat production planning in PT Wiska) . Departemen

Manajemen. Bogor: Institut Pertanian Bogor.

Pratama, T. I. (1999). Metode Peramalan Memakai Jaringan Saraf Buatan dengan Cara Backpropagation (forecasting using

backpropagation artificial neural network) . Jurnal Teknologi Industri , III (2), 109-116.

Pujihastuti, E. (2008). Analisis Kebijakan Perusahaan dalam Pengendalian Persediaan Bahan Baku di PT. X (Policy analysis in

controlling inventory). Program Sarjana Ekstensi Manajemen Agribisnis. Bogor: Institut Pertanian Bogor.

Sari, S. P. (2010). Pengoptimalan Persediaan Bahan Baku Kacang Tanah Menggunakan Metode EOQ (Economic Order Quantity) di PT.

Dua Kelinci Pati (Optimalization of butter as inventory using EOQ in PT dua Kelinci Pati). Universitas Sebelas Maret, Fakultas

Pertanian. Surakarta: Universitas Sebelas Maret.

Tama, B. A. (2010). Marketing Strategic Using Association Rule on The CRM Perspective (Penetapan Strategi Penjualan Menggunakan

Association Rules dalam Konteks CRM). Jurnal Generic , 5 (1), 36. Palembang

Yusuf, Y., Pratikto, F. R., & Gerry. (2006). Penerapan Data Mining Dalam Penentuan Aturan Asosiasi Antar Jenis Item (items association

using data mining). Seminar Nasional Aplikasi Teknologi Informasi (SNATI), (hal. 53-56). Yogyakarta.

quotesdbs_dbs17.pdfusesText_23