氏名　SAIFUL ANWAR
学位の種類　博士（工学）
学位記番号　博甲第610号
学位授与の日付　平成24年3月26日
学位論文題目　Prototype Development of A Web-based Decision Support System for Islamic Bank's Depositors　（イスラム銀行預金者向け意思決定支援システムのプロトタイプ開発）
論文審査委員
　主査　教授　三上　喜貴
　副査　教授　淺井　達雄
　副査　教授　李　志東
　副査　准教授　マーラシンハ チャンドラジット アーシュボーダ
　副査　名古屋工業大学教授　渡辺　研司

• 論文目次
• 論文要旨
• 論文審査の結果の要旨

［平成23（2011）年度博士論文題名一覧］ ［博士論文題名一覧］に戻る．

TABLE OF CONTENTS　page
ACKNOWLEDGEMENT　p.i
TABEL OF CONTENT　p.iii
LIST OF FIGURES AND TABLES　p.vii
ABSTRACT　p.xiii
CHAPTER 1.　INTRODUCTION
　1.1　General theory of saving motive　p.1
　1.2　Investment-saving motive of Islamic bank's depositors　p.4
　1.3　Islamic compliant saving product　p.6
　1.4　Decision support system （DSS） issue related with profit motive　p.7
　1.5　Prediction algorithm issue in DSS application　p.8
　1.6　Motivation of research　p.11
　1.7　Research problem　p.15
　1.8　Research question　p.16
　1.9　Theoretical positioning if the research problem　p.19
　1.10　Originality of the research　p.19
　1.11　Thesis outline and contents　p.23
CHAPTER 2.　LITERATURE REVIEW
　2.1　Introduction　p.27
　2.2　Banking operation　p.30
　2.2.1　Accepting deposit　p.32
　2.2.2　Channeling the funds　p.33
　2.2.3　Providing agency activities　p.35
　2.2.4　How do banks set the interest rate　p.37
　2.2.4.1　Lending and deposit interest rate　p.38
　2.2.4.2　Interest rate pricing and macroeconomic condition　p.41
　2.3　Islamic banking industry　p.43
　2.3.1　Basic theory of Islamic banking　p.43
　2.3.2　The operational of Islamic bank　p.44
　2.3.3　The products of Islamic bank　p.44
　2.3.4　Indonesian Islamic bank　p.46
　2.4　Decision support system　p.48
　2.4.1　Model management subsystem　p.50
　2.4.2　Prediction algorithm　p.52
　2.4.2.1　Multipule liner regression（MLR）　p.53
　2.4.2.2　Generalized autoregressive conditional heteroskedasticity （GARCH）　p.53
　2.4.2.3　Artificial neural networks （ANNs）　p.55
　2.4.2.4　Support vector machines （SVMs）　p.58
CHAPTER 3.　RESEARCH METHODOLOGY AND DATA PREPARATION
　3.1　Introduction　p.63
　3.2　General process of prototype development　p.64
　3.3　Step 1: Independent variables selection　p.70
　3.4　Step 2: Prediction algorithm selection　p.75
　3.5　Step 3: Testing the selected algorithm　p.81
CHAPTER 4.　RESULTS AND DISCUSSION
　4.1　Introduction　p.85
　4.2　Step 1: Independent variables selection　p.88
　4.2.1　Data analysis　p.88
　4.2.2　Data pre-processing　p.88
　4.2.3　Designing the networks architecture　p.88
　4.2.4　Training the networks　p.89
　4.2.5　Testing the robustness of networks　p.90
　4.3　Step 2: Prediction algorithm selection　p.94
　4.3.1　GARCH model　p.94
　4.3.1.1　The results　p.97
　4.3.2　ANNs model　p.100
　4.3.2.1　The results　p.102
　4.3.3　SVMs model　p.105
　4.3.3.1　The results　p.106
　4.4　Comparison of accuracy rate prediction　p.110
　4.5　Conclusion　p.114
CHAPTER 5.　THE PROTOTYPE AND ITS IMPLEMENTATION
　5.1　Introduction　p.117
　5.2　The accuracy rate performance　p.120
　5.3　Validation testing on accuracy performance in various phase of economic　p.124
　5.3.1　First testing: During stable economic condion　p.127
　5.3.2　Second testing: During stable economic condition　p.128
　5.3.3　Third testing: After global financial crisis period　p.129
　5.4　Conclusion　p.131
CHAPTER 6.　CONCLUSIONS, IMPACT AND SUGGESTIONS FOR FURTHER RESEARCH
　6.1　Introduction　p.133
　6.2　Adresing the research problems and the questions　p.134
　6.3　The function of the proposed DSS　p.135
　6.4　The impact of the proposed DSS　p.136
　6.4.1　The theoretical impacts　p.136
　6.4.2　The pratical impacts　p.137
　6.5　The proposed system limitaions　p.138
　6.6　Suggestiond for further research　p.138
　6.6.1　The extendion function of proposed DSS　p.138
　6.6.2　The business model　p.139
REFERENCES　p.141

This research proposes a prototype of decision support system （DSS） for depositors of Islamic banks using World Wide Web as an interface. The system supports depositors profit motive in dealing with investment based deposit product namely mudharabah time deposit. The underlying theoretical assumption could be said as follows: “The deposit return of Islamic banks should be treated as pure investment instrument that needs specific supporting system”. Through that new assumption, the proposed system contributes to the theoretical field of profit motive behavior of Islamic bank’s depositors by implementing it into information system application. Shortly, the research investigates the pattern of deposit return as proxy of profitability performance through a set of exogenous variables. It follows macro stress-testing model applied in banking industry which is used as early warning of financial performance, estimating bank’s profitability using only macroeconomic variables as input.
Specifically, the research investigation is divided into three consecutive parts. Part one is to perform sensitivity analysis to select some variables from a set of independent variables which give significant impact in influencing dependent variable. The contribution of this part into Islamic bank and finance literature is to use different approach modifying previous investigation in selecting macroeconomic variables as determinants of Islamic bank’s performance through two underlying channels, borrower and depositor. The borrower channel explains that volatility of specific macroeconomic variables affects borrower’s business which subsequently affects Islamic bank’s profit through two factors: number of borrowers who end up in default because failure to pay and the increasing or decreasing of shareable profit generated from good performing borrower. The variables used from this channel are exchange rate （EXCH）, prime rate （BIRT）, money supply （M1）, inflation rate （INFR）, and oil prices （OIL）. Meanwhile, depositor channel explains that the volatility of deposit fund affects number of money channelized to borrowers which influences Islamic bank’s performance later on. The background of addressing this channel in determining Islamic bank’s profitability is because there is no breakage fee attached to time deposit product. This allows depositors to withdraw the fund anytime responding to other investment instrument such as gold price （GOLD）, stock index （STIN）, and market interest rate （INTR）. Using artificial neural networks （ANNs） model, the investigation reveals that OIL and GOLD variables should be omitted due to their very small contribution.

Part two is to select prediction algorithm that will be employed in the proposed system as prediction machine. This step is considered to be a new contribution to Islamic banking research literature by which confirming the predicting ability of widely used algorithms, generalized autoregressive conditional heteroskedastic （GARCH） and ANNs, with support vector machines （SVMs） as a newest model in machine learning technique. Each model has its own uniqueness which are interesting to be explored. GARCH model which is introduced by Engle, a Nobel laureate, exploits non constant volatility feature of financial time series data the so-called volatility clustering in which the data tends to exhibit period of high and low volatility. Moreover, ANNs model is a part of machine learning technique which tries to simulate the way of learning of human brain. Furthermore, the SVMs model is developed to overcome the overfitting problem of ANNs model. The model yields a linear model constructed in a higher dimensional input space which represents the non linear decision boundary in the original space. Using EXCH, INFR, M1, BIRT, INTR and BIRT, the investigation found that ANNs model is the best suited algorithm to predict deposit return based on accuracy rate and statistical parameters.
In the last part, the ANNs algorithm is embedded into the system to perform real case prediction. The system is employed to predict time deposit product of the biggest Islamic bank in Indonesia namely Bank Syariah Mandiri. It successfully completes the task achieving 99.99% accuracy rate. The additional investigation is then conducted to check the performance in various phases of economic cycles. The investigation discovers that the algorithm performs very well, thus giving satisfying result.
While the system is limited only to serve prediction tasks in Indonesian Islamic banking industry, thus the replication in different countries may need modification of parameters or algorithms. The full implementation of this system may bring business opportunity as commercial website such as Google advisor that delivers interest rate comparison of all banks in United States side by side. Therefore, a specific business model will be created to serve eight million Islamic bank’s depositors in particular and to promote Islamic deposit products to 42 million of conventional bank’s depositors.

本論文は、”Prototype Development of a web-based decision support system for Islamic bank’s depositors”（イスラム銀行預金者向け意思決定支援システムのプロトタイプ開発）」と題し、7章より構成されている。
まず、この特殊な主題についての背景を説明する。イスラム銀行とはイスラムの律法に従って経営される銀行であり、通常の銀行と異なり、預金者はあらかじめ定められた金利ではなく、事後的な利益配分という形で資金運用益の配分を受ける。従来、イスラム銀行預金者は宗教的動機付けにより預金を行うと考えられてきたが、近年の研究により、通常の預金者と同様に利潤動機に基づく預金行動をとっていることが明らかとなった。しかしながら預金者にとっては事後的に決まる利益配分額を予想することができず、この情報不足がイスラム銀行にとって通常銀行との預金獲得競争上の障害となっていた。本論文はこの障害を取り除くべく、イスラム銀行の利益予測を行い、預金者の意思決定を支援するシステムのプロトタイプを開発しようという目的をもって行われた研究の成果である。
第1章”Introduction”で研究の動機、目的、計画等について述べる。
第2章”Literatue Review”は、イスラム銀行のオペレーション、預金者行動に関する経済理論的研究、意思決定支援システムに関する技術的文献などの各領域について先行研究のレビューを行う。
第3章”Research Methodology and Data Preparation”は、イスラム銀行の利益予想の方法論について、説明変数の選択、アルゴリズムの選択、アルゴリズムの比較検討の3段階で述べる。
第4章”Results and Discussion”は、イスラム銀行の利益予想モデルとして、ニューラル・ネットワーク（ANN）、時系列統計モデル、統計的機械学習モデルを比較検討した結果について述べる。
第5章”The Prototype and Its Implementation”は、以上の比較検討の結果最適と判断されたANNモデルに基づく利益予測の詳細およびウェブ上で稼動する預金者支援プロトタイプシステムについて述べる。
第6章”Conclusions, Impact and Suggestions for Further Research”は、結論及び今後の研究課題について述べる。
本研究は、イスラム銀行預金者のための意思決定支援システムに関する研究として世界的に先駆的なものであり、既に、インドネシア中央銀行主催のシンポジウムで優秀発表賞を受賞するなど、イスラム銀行関係者から高い評価と注目を得ている。よって、本論文は工学上及び工業上貢献するところが大きく、博士（工学）の学位論文として十分な価値を有するものと認める。