Minggu, 20 November 2011
Selasa, 15 November 2011
MENGINTEGRASIKAN MODEL PERILAKU KONSUMEN MEMBELI KE PENGEMBANGAN AGEN BERBASIS SISTEM E-COMMERCE
Integrasi Guttman Perilaku Konsumen Membeli model (CBB) ke dalam MAS-Common KADS metodologi untuk mengembangkan sistem multi agen. Electronic commerce (E-Commerce) adalah suatu teknologi yang sedang berkembang pesat di Dunia Wide Web, dan telah menjadi pelengkap yang terdapat dalam kegiatan usaha perusahaan dan individu. Teknologi agen dianggap sebagai tingkat berikutnya yang lebih tinggi abstraksi dalam model solusi berbasis e-commerce aplikasi (Papazoglou 2001). Analisis dan desain dari suatu yang direkomendasikan berbasis agen sistem untuk membeli produk dari toko-toko virtual yang berbeda, menggunakan metodologi MAS-CommonKADS (Iglesias et al. 1998a). Metodologi ini ditingkatkan dengan menggabungkan beberapa aspek dari pembelian konsumen
model perilaku Guttman (Guttman et al 1998.),
MODEL PERILAKU PEMBELIAN KONSUMEN
Sebuah model perilaku konsumen membeli (CBB) biasanya terdiri dari tindakan dan keputusan yang terlibat dalam membeli dan menggunakan barang dan jasa. Model Khas berfokus pada pasar ritel meskipun mereka dapat berhubungan dengan
bisnis-ke-bisnis dan konsumen ke konsumen pasar (Andreasen 1965; Howard dan Sheth
1969; Hawkins et al. 1980; Engel dan Blackwell 1982).
Model yang diperkenalkan oleh Guttman dalam (Guttman et al. 1998), yang memperpanjang model khas dengan konsep-konsep dari penelitian perangkat lunak agenuntuk
mengakomodasi pertumbuhan pasar elektronik. Model Guttman mengusulkan enam tahap yang dapat memandu pembelian perilaku konsumen dalam konteks pasar elektronik, yaitu :
1)Perlu Identifikasi. Tahap ini berkaitan dengan identifikasi kebutuhan konsumen. dalam hal ini
panggung, konsumen dapat distimulasi melalui informasi produk.
2) Produk brokering. Tahap ini adalah di mana konsumen menentukan apa yang harus membeli. Hal ini terjadi setelahkebutuhan telah telah diidentifikasi. Keputusan ini dibuat setelah kritis evaluasi informasi produk diambil. Hasil dari tahap ini adalah sebuah "set membangkitkan" produk.
3)brokering Merchant. Tahap ini menggabungkan "membangkitkan set" dari tahap sebelumnya dengan pedagang-spesifik informasi untuk membantu menentukan siapa yang membeli.
4) Negosiasi Tahap ini berhubungan dengan bagaimana menentukan persyaratan transaksi.
5)Pembelian dan Pengiriman. Tahap ini mencakup pembelian dan pengiriman produk.
6) Produk Layanan dan Evaluasi. tahap ini melibatkan layanan produk, layanan pelanggan, dan evaluasi kepuasan dari pembelian keseluruhan pengalaman dan keputusan.
Gutmman menunjukkan bahwa tahapan ini terjadi di beberapa agen sistem untuk e-commerce, khususnya Produk Brokering, Merchant brokering dan Negosiasi tahap (Guttman et al. 1998). Meskipun seperti berbagai perdagangan elektronik sistem termasuk tahap CBB, tidak ada referensi untuk cara tahap CBB mempengaruhi desain multi agen sistem untuk perdagangan elektronik.
METODOLOGI MAS-CommonKADS
Tujuan analisis multi agen dan metodologi desain, memperluas CommonKADS (Schreiber et al. 1999)
metode desain dengan menggabungkan teknik dari metodologi objectoriented dan rekayasa protokol. Terdapat tujuh model yang mencakup aspek utama pengembangan multi agen sistem:
1)Model Agen menentukan karakteristik agen
2) Model tugas menjelaskan tugas-tugas yang dilaksanakan para agen
3) Model keahlian mendefinisikan pengetahuan yang dibutuhkan oleh agen untuk mencapai tujuan mereka.
4) Model organisasi menggambarkan social organisasi masyarakat .
5)Model koordinasi menggambarkan percakapan antara agen.
6)Rincian model komunikasi interaksi agen humansoftware.
7) Model desain mencakup, di samping tindakan khas dari tahap desain.
.
Penerapan metodologi ini terdiri dalam mengembangkan model yang berbeda. Ini telah
berhasil diterapkan untuk optimasi system untuk aplikasi rollmill(Iglesias et al. 1998b) dan otomatisasi asisten perjalanan (Arenas dan Barrera-Sanabria 2002).
MENGINTGRASIKAN MODEL Guttman CBB DENGAN MAS-CommonKADS
penerapan MAS-CommonKADS ke desain sistem direkomendasikan untuk membeli buku-buku dari toko buku virtual yang berbeda. Pengguna dapat mencari Buku dengan memasukkan nama penulis atau kata kunci dari judul buku. Maka sistem menampilkan opsi tersedia dari toko buku yang telah ditentukan, yang menunjukkan termurah yang tersedia pilihan. Akhirnya, pengguna dapat memilih pilihan dan membeli produk yang diinginkan.
Agen Model
Model agen menentukan karakteristik agen dan memainkan peran titik acuan untuk model lainnyal. Agen didefinisikan sebagai entitas apapun manusia atau
perangkat lunak yang mampu melaksanakan suatu kegiatan(Iglesias et
al. 1998a). Para agen dapat diidentifikasi sebagai berikut::
• Klien Agen. Manusia agen yang berinteraksi dengan sistem untuk mendapatkan informasi tentang buku-buku yang tersedia di toko-toko buku yang berbeda.
• Recommender Agen. Perangkat lunak agen yang menentukan pilihan belanja terbaik.
• Merchant Agen. Perangkat lunak agen (satu untuk setiap berpartisipasi toko buku virtual) .
• Pembeli Agen. Perangkat lunak agen yang membeli produk menggunakan informasi yang diberikan oleh klien.
Model Tugas.
Model tugas menjelaskan tugas-tugas yang harus agen laksanakan.
Model Koordinasi.
Model koordinasi menunjukkan dinamika hubungan antara agen-agen. Ini dimulai dengan identifikasi percakapan antara agen, di mana
kasus penggunaan bermain lagi peran penting.
Model Komunikasi.
Model komunikasi mencakup interaksi antara manusia agen dan agen lainnya. Kami menggunakan template mirip dengan model koordinasi.
Model Keahlian.
Model keahlian menggambarkan kemampuan penalaran agen yang dibutuhkan untuk melaksanakan tugas-tugas merekadan mencapai tujuan mereka.
Model Organisasi.
Studi organisasi adalah alat untuk identifikasi dampak pada multi agen yang
sistem ketika diinstal. Model organisasi menentukan hubungan struktural antara agen-agen
dengan mewakili diagram kelas baik agen dan mendefinisikan contoh agen. Dibandingkan dengan berorientasi objek paradigma,
,
model perilaku Guttman (Guttman et al 1998.),
MODEL PERILAKU PEMBELIAN KONSUMEN
Sebuah model perilaku konsumen membeli (CBB) biasanya terdiri dari tindakan dan keputusan yang terlibat dalam membeli dan menggunakan barang dan jasa. Model Khas berfokus pada pasar ritel meskipun mereka dapat berhubungan dengan
bisnis-ke-bisnis dan konsumen ke konsumen pasar (Andreasen 1965; Howard dan Sheth
1969; Hawkins et al. 1980; Engel dan Blackwell 1982).
Model yang diperkenalkan oleh Guttman dalam (Guttman et al. 1998), yang memperpanjang model khas dengan konsep-konsep dari penelitian perangkat lunak agenuntuk
mengakomodasi pertumbuhan pasar elektronik. Model Guttman mengusulkan enam tahap yang dapat memandu pembelian perilaku konsumen dalam konteks pasar elektronik, yaitu :
1)Perlu Identifikasi. Tahap ini berkaitan dengan identifikasi kebutuhan konsumen. dalam hal ini
panggung, konsumen dapat distimulasi melalui informasi produk.
2) Produk brokering. Tahap ini adalah di mana konsumen menentukan apa yang harus membeli. Hal ini terjadi setelahkebutuhan telah telah diidentifikasi. Keputusan ini dibuat setelah kritis evaluasi informasi produk diambil. Hasil dari tahap ini adalah sebuah "set membangkitkan" produk.
3)brokering Merchant. Tahap ini menggabungkan "membangkitkan set" dari tahap sebelumnya dengan pedagang-spesifik informasi untuk membantu menentukan siapa yang membeli.
4) Negosiasi Tahap ini berhubungan dengan bagaimana menentukan persyaratan transaksi.
5)Pembelian dan Pengiriman. Tahap ini mencakup pembelian dan pengiriman produk.
6) Produk Layanan dan Evaluasi. tahap ini melibatkan layanan produk, layanan pelanggan, dan evaluasi kepuasan dari pembelian keseluruhan pengalaman dan keputusan.
Gutmman menunjukkan bahwa tahapan ini terjadi di beberapa agen sistem untuk e-commerce, khususnya Produk Brokering, Merchant brokering dan Negosiasi tahap (Guttman et al. 1998). Meskipun seperti berbagai perdagangan elektronik sistem termasuk tahap CBB, tidak ada referensi untuk cara tahap CBB mempengaruhi desain multi agen sistem untuk perdagangan elektronik.
METODOLOGI MAS-CommonKADS
Tujuan analisis multi agen dan metodologi desain, memperluas CommonKADS (Schreiber et al. 1999)
metode desain dengan menggabungkan teknik dari metodologi objectoriented dan rekayasa protokol. Terdapat tujuh model yang mencakup aspek utama pengembangan multi agen sistem:
1)Model Agen menentukan karakteristik agen
2) Model tugas menjelaskan tugas-tugas yang dilaksanakan para agen
3) Model keahlian mendefinisikan pengetahuan yang dibutuhkan oleh agen untuk mencapai tujuan mereka.
4) Model organisasi menggambarkan social organisasi masyarakat .
5)Model koordinasi menggambarkan percakapan antara agen.
6)Rincian model komunikasi interaksi agen humansoftware.
7) Model desain mencakup, di samping tindakan khas dari tahap desain.
.
Penerapan metodologi ini terdiri dalam mengembangkan model yang berbeda. Ini telah
berhasil diterapkan untuk optimasi system untuk aplikasi rollmill(Iglesias et al. 1998b) dan otomatisasi asisten perjalanan (Arenas dan Barrera-Sanabria 2002).
MENGINTGRASIKAN MODEL Guttman CBB DENGAN MAS-CommonKADS
penerapan MAS-CommonKADS ke desain sistem direkomendasikan untuk membeli buku-buku dari toko buku virtual yang berbeda. Pengguna dapat mencari Buku dengan memasukkan nama penulis atau kata kunci dari judul buku. Maka sistem menampilkan opsi tersedia dari toko buku yang telah ditentukan, yang menunjukkan termurah yang tersedia pilihan. Akhirnya, pengguna dapat memilih pilihan dan membeli produk yang diinginkan.
Agen Model
Model agen menentukan karakteristik agen dan memainkan peran titik acuan untuk model lainnyal. Agen didefinisikan sebagai entitas apapun manusia atau
perangkat lunak yang mampu melaksanakan suatu kegiatan(Iglesias et
al. 1998a). Para agen dapat diidentifikasi sebagai berikut::
• Klien Agen. Manusia agen yang berinteraksi dengan sistem untuk mendapatkan informasi tentang buku-buku yang tersedia di toko-toko buku yang berbeda.
• Recommender Agen. Perangkat lunak agen yang menentukan pilihan belanja terbaik.
• Merchant Agen. Perangkat lunak agen (satu untuk setiap berpartisipasi toko buku virtual) .
• Pembeli Agen. Perangkat lunak agen yang membeli produk menggunakan informasi yang diberikan oleh klien.
Model Tugas.
Model tugas menjelaskan tugas-tugas yang harus agen laksanakan.
Model Koordinasi.
Model koordinasi menunjukkan dinamika hubungan antara agen-agen. Ini dimulai dengan identifikasi percakapan antara agen, di mana
kasus penggunaan bermain lagi peran penting.
Model Komunikasi.
Model komunikasi mencakup interaksi antara manusia agen dan agen lainnya. Kami menggunakan template mirip dengan model koordinasi.
Model Keahlian.
Model keahlian menggambarkan kemampuan penalaran agen yang dibutuhkan untuk melaksanakan tugas-tugas merekadan mencapai tujuan mereka.
Model Organisasi.
Studi organisasi adalah alat untuk identifikasi dampak pada multi agen yang
sistem ketika diinstal. Model organisasi menentukan hubungan struktural antara agen-agen
dengan mewakili diagram kelas baik agen dan mendefinisikan contoh agen. Dibandingkan dengan berorientasi objek paradigma,
,
Senin, 14 November 2011
INTEGRATING A CONSUMER BUYING BEHAVIOUR MODEL INTO THE DEVELOPMENT OF AGENT-BASED E-COMMERCE SYSTEMS
INTEGRATING A CONSUMER BUYING BEHAVIOUR MODEL INTO
THE DEVELOPMENT OF AGENT-BASED E-COMMERCE SYSTEMS
ABSTRACT
This paper illustrates the integration of Guttman’s
Consumer Buying Behaviour model (CBB) into the
MAS-CommonKADS methodology for developing
multiagent systems. We develop each model included in
this methodology for the construction of a recommender
system for buying products from different virtual shops;
and show the integration of the stages of the CBB model
within the MAS-CommonKADS models. We also
describe the implementation of the system using the
agent-building tool ZEUS.
1. INTRODUCTION
Electronic commerce is growing rapidly in the World
Wide Web, and has become a complement to the usual
business activities of corporations and individuals.
Agent technology is considered as the next higher level
abstraction in model-based solutions to e-commerce
applications (Papazoglou 2001). This paper illustrates
the analysis and design of an agent-based recommender
system for buying products from different virtual shops,
using the MAS-CommonKADS methodology (Iglesias
et al. 1998a). The methodology is improved by
incorporating some aspects of the consumer buying
behaviour model of Guttman (Guttman et al. 1998),
when designing e-commerce systems.
The following section describes the consumer buying
behaviour model of Guttman. Section 3 illustrates the
MAS-CommonKADS methodology. Section 4 presents
the analysis and design of the recommender system
using the MAS-CommonKADS methodology,
highlighting how the consumer buying behaviour model
influenced some steps in the methodology. Next, section
5 describes the implementation of a system prototype
using ZEUS (Nwana et al. 1999). Section 6 comparesour work with previous research. Finally, section 7
gathers some concluding remarks and discusses possible
extensions to our work
2. CONSUMER BUYING BEHAVIOUR MODEL
A model of consumer buying behaviour (CBB) usually
comprises the actions and decisions involved in buying
and using goods and services. Typical models focus on
the retail market although they can be related to
business-to-business and consumer-to-consumer
markets as well (Andreasen 1965; Howard and Sheth
1969; Hawkins et al. 1980; Engel and Blackwell 1982).
In this paper we use the model introduced by Guttman
in (Guttman et al. 1998), which extends typical models
with concepts from software agents research in order to
accommodate the growing electronic markets.
The Guttman model proposes six stages that may guide
the consumer buying behaviour in the context of
electronic markets :
• Need Identification. This stage deals with the
identification of consumer’s needs. Within this
stage, the consumer can be stimulated through
product information.
• Product Brokering. This stage is where consumers
determine what to buy. This occurs after a need has
been identified. The decision is made after a critical
evaluation of the retrieved product information. The
result of this stage is an "evoked set" of products.
• Merchant Brokering. This stage combines the
"evoked set" from the previous stage with
merchant-specific information to help determine
who to buy from.
• Negotiation This stage is connected with how to
determine the terms of the transaction.
• Purchase and Delivery. This stage includes the
purchase and delivery of a product.
• Product Service and Evaluation. This stage
involves product service, customer service, and an
evaluation of the satisfactions of the overall buying
experience and decision.
Gutmman shows that these stages occur in several agent
systems for e-commerce, particularly the Product
Brokering, Merchant Brokering and Negotiation stages
(Guttman et al. 1998). For instance, Firefly
(Shardanand and Maes 1995) is a system that helps
consumers find products, as part of the Product
Brokering stage; Jango (Doorenbos et al. 1997) is a
shopping agent that includes price comparisons from
different merchant Web sites, corresponding to the
Merchant Brokering stage; Kasbah (Chavez et al. 1997)
is a multiagent system for consumer-to-consumer
electronic commerce where users create buying and
selling agents to help transact products, automating
much of the Merchant Brokering and Negotiation
stages.
Despite such a wide range of electronic commerce
systems including CBB stages, there is no reference to
the way CBB stages influence the design of multiagent
systems for electronic commerce.
3. THE MAS-CommonKADS METHODOLOGY
MAS-CommonKADS (Iglesias et al. 1998a) is a general
purpose multiagent analysis and design methodology, it
extends the CommonKADS (Schreiber et al. 1999)
design method by combining techniques from objectoriented methodologies and protocol engineering. The
methodology is centred around seven models that cover
the main aspects of the development of multiagent
systems:
• The agent model specifies agents characteristics
such as reasoning capabilities, sensor/effectors,
services, agent groups and hierarchies.
• The task model describes the tasks that the agents
can carry out, such as goals, decomposition,
problem-solving methods, etc.
• The expertise model defines the knowledge needed
by the agents to achieve their goals.
• The organisation model describes the social
organisation of the agent society.
• The coordination model illustrates the conversation
between agents.
• The communication model details the humansoftware agent interactions.
• The design model includes, in addition to the
typical action of the design phase (Pressman 2001),
the design of relevant aspects of the agent network,
selecting the most suitable agent architecture and
the agent development platform.
The application of the methodology consists in
developing the different models. It has been
successfully applied to systems for optimisation of rollmill applications (Iglesias et al. 1998b) and automation
of travel assistants (Arenas and Barrera-Sanabria 2002)
among others.
4. INTEGRATING GUTTMAN CBB MODEL
WITH MAS-CommonKADS
In this section we apply MAS-CommonKADS to the
design of a recommender system for buying books from
different virtual bookshops. The user can look for a
book by entering the author’s name or keywords from
the book title. Then the system shows the options
available from predetermined bookshops, indicating the
cheapest available option. Finally, the user can choose
an option and buy the desired product.
4.1. Conceptualisation
This initial phase aims to carry out an elicitation task in
order to obtain a general description of the problem,
following a user-centred approach based on use cases
(Rambaugh et al. 1999). In this approach, an actor
represents a role played by a person, a piece of
hardware or another system that interacts with our
system. A use case corresponds to a description of the
sequence of actions needed to produce an observable
result useful for an actor.
The CBB model has been useful to identify the actors
and uses cases. The Need Identification stage can be
carried out by including facilities for obtaining client
preferences and informing them about new products in a
proactive way. These facilities are not fully incorporated
in the current version of our system, we have only
included facilities for obtaining basic information about
the actor Client, facilities that are described in the use
case Client Information. In order to confirm the
Product Brokering stage, we include the following
actors: Merchant Agents (one for each participating
bookshop), and a Recommender Agent. A Client makes
a Product Request to the Recommender Agent, who
Requests Merchant Information to each Merchant Agent
in order to obtain the evoked set of products. In the
Merchant Brokering stage the Recommender Agent does
Display Recommendations and Select Best Option.
Traditional retail markets usually do not leave room for
negotiation of price and other aspects of the
transactions. However, a negotiation is central to
markets such as stocks, fine arts, etc. We have not yet
included aspects of the Negotiation stage in our system.
In order to agree with the Purchase and Delivery stage,
we include a Buyer Agent that buys products on behalf
of a client. The Client agent makes a Buy Request, and
the Buyer Agent does Verify Information and performs
the Buy. Finally, the Produce Service and Evaluation
stage can be achieved by means of the evaluation forms
that the Merchant Agent requests to the Client Agent;
this facility will be included in future versions of the
system.
Table 1 summarises the actors and uses cases that we
have identified, as well as the corresponding CBB
stages.
The outcome of the conceptualisation phase is a
description of the different actors and their use cases. As
a way of illustration, Table 2 describes the textual
template of use case Request Merchant Information for
actor Recommender Agent.
Table 2. Textual Template of Use Case Request
Merchant Information for Actor Recommender
Agent
Use Case Request Merchant Information
Summary. The Recommender Agent takes the
information about a book given by the Client and
requests the different Merchant Agents for
additional information (title, price, availability),
giving a list of recommendations.
Actors. Recommender Agent and Merchant Agent
Precondition. Book information and list of
merchant agents.
Postcondition. List of recommendations
Exceptions. No book information; empty list of
merchant agents.
4.2. Analysis
This phase results in the requirement specification of the
system through the development of the first six models
of the methodology. The Agent Model.
The agent model specifies the characteristics of an agent
and plays the role of a reference point for the other
models. An agent is defined as any entity - human or
software - capable of carrying out an activity (Iglesias et
al. 1998a). The identification of agents is based on the
actors and use cases generated in the previous phase of
conceptualisation. The agents we have identified are:
• Client Agent. Human agent who interacts with the
system to obtain information about books available
in the different bookshops.
• Recommender Agent. Software agent that
determines the best shopping option.
• Merchant Agent. Software agent (one for each
participating virtual bookshop) that acts as an interface to the bookshop by offering information
about its available books.
• Buyer Agent. Software agent that buys products
using information provided by the client.
Although we have here an one-to-one association
between actors and agents, this is not always the case.
We can either merge several actors into an agent or
decompose an actor into two o more agents. The
outcome of this model is a set of textual templates (one
for each agent), that shows information such as
description of the agent, parameters, services, etc. For
instance, Table 3 presents the template of the
Recommender Agent.
Table 3. Textual Template of the Recommender Agent
Agent Recommender Agent
Name. Recommender
Type. Software agent
Role. Information provider
Location. Inside the agent society
Description. This agent determines the best
shopping option according to the request made by a
client and the information provided by the
merchants.
Objective. Provides the client with additional
information about the requested book.
Exceptions. Lack of information about a book;
absence of available information in the merchants
about the requested book.
Input Parameters. Book keywords and
information provided by the Merchant Agents.
Output Parameters. List of availability of the
requested book in each bookshop, including the
price and delivering time.
Expertise. This agent must know the profile of the
client who requests a book, in order to apply
heuristics for the recommendation process. It also
stores information about given recommendations so
that the learning process may start.
Communication. Client Agent.
Coordination. Merchant Agent
The Task Model.
The task model describes the tasks that the agent can
carry out. Tasks are decomposed according to a topdown approach. Following (Arenas and BarreraSanabria 2002), we use UML activity diagrams
(Rambaugh et al. 1999) to represent the activity flow of
the tasks. Figure 1 shows the activity diagram for the
Recommender Agent.
Each task is described using a textual template that
includes its name, a short description, input and output
ingredients and required capabilities of the performers.
This documentation is used to support the maintenance
and management of changes in the organisation and
feasibility assessment (Iglesias et al. 1998a). Table 4
shows the textual template of task Determine Best
Shopping Option.
Table 4. Textual Template of Task Determine Best
Shopping Option.
Task Determine Best Shopping Option
Objective. Determine best shopping option from
all the options generated by the virtual bookshops.
Description. Organises the list of products
generated by the Infer Information task from the
cheapest option to the more expensive one. The
cheapest option corresponds to the best shopping
option.
Input Ingredient. List of available products.
Output Ingredient. Best shopping option.
Constraints. None
Exceptions. Absence of products.
The Coordination Model.
The coordination model shows the dynamic
relationships between the agents. It begins with the
identification of the conversation between agents, where
use cases play again an important role. At this level,
every conversation just consists of one single interaction
and the possible answer, which are described by means
of templates as illustrated in Table 5.
Table 5. Textual Template of Conversation Get
Shopping Options
Conversation Get Shopping Options
Type. Getting information.
Objective. Get available books in the virtual
bookshops according to the information provided
by the client.
Agents. Recommender Agent, Merchant Agents
and Client.
Beginner. Recommender Agent.
Service. List of bookshops where the requested
book is available.
Description. The Recommender Agent request for
information about the book to the different
Merchant Agents. The Recommender Agent has a
limited time to get information from other agents.
In case there is a timeout, it sends immediately a
message to the Client notifying it was not possible
to finish the search.
Precondition. Book keywords.
Postcondition. List of bookshops where the book
is available, including price, delivering time and a
short additional information.
Ending condition. Process aborted by the client or
absence of Merchant Agent information.
Next, we model the data exchanged in each interaction
by specifying speech acts and synchronisation, and
collect all this information in the form of sequence
diagrams. In Figure 2, we show the interactions of the
Recommender Agent.
The Communication Model.
The communication model includes interaction between
human agents and other agents. We use templates
similar to those of the coordination model, but taking
into consideration human factors such as facilities for
understanding the recommendation given by the system.
In our system, the communication model includes
interaction between the Client and the Recommender
and the Buyer Agents.
The Expertise Model.
The expertise model describes the reasoning capabilities
of the agents needed to carry out their tasks and achieve
their goals. It consists of two submodels: the
application knowledge and the problem solving
knowledge.
The development of the application knowledge includes
to represent the declarative knowledge of the problem
(domain knowledge), modelled as concepts, properties,
expressions and relationships using the Conceptual
Modelling Language (Schreiber et al. 1999) or graphical
notation of the Object Model of OMT (Milsted 1995);
and to represent the inference steps performed for
solving a task and the order of the inference structures
(inference and task knowledge). In our system, the only
agent that uses knowledge is the Recommender Agent.
We have identified concepts such as books, merchants,
promotions, and properties such as price, book
keywords, etc. These concepts are organised in domain
models that describe particular relationships between
themselves. For instance, we have developed a casual
model of dates of promotions for particular merchants.
The Recommender Agent performs also the generic task
of inferring the best shopping option, including
knowledge about availability of the book in the
bookshops and its price.
The problem solving knowledge includes the
specification of a problem solving method. We use
methods such as assessment that generates schemes for
organising the knowledge (Schreiber et al. 1999). This
method includes the representation of the relevant
knowledge (for instance, special discount for lectures),
the definition of norms of integrity, and its definition in
a representation language.
The Organisation Model.
The study of the organisation is a tool for the
identification of possible impact on the multiagent
system when installed. The organisation model
specifies the structural relationships between the agents
by representing both class agent diagrams and defining
instances of agents. In comparison to the object-oriented
paradigm, the agent-instance information is more
relevant than the class agent diagrams (Iglesias et al.
1998a). Figure 3 shows the class agent diagram of our
system and Table 6 the agent instances.
4.3. Design
In this phase the previous models are used as a basis for
the design model, which consists of the agent network,
the agent design and the platform submodels.
The agent network design model describes the network
facilities (naming services, security, encryption),
knowledge facilities (ontology servers, knowledge
representation translators) and the coordination facilities
(protocol servers, group management facilities, resource
allocation) within the target system. In our prototype
system, we do not use any facilitator agent, so that this
model is not defined.
The agent design model determines the most suitable
architecture for each agent. The method suggests a
generic agent architecture that consists of a usercommunication unit for the agent-user interaction (from
the communication model), an agent communication
unit for the interaction among agents (from the
coordination model) and a deliberation and reaction unit
for the reasoning of the agent (from the agent, expertise
and task models).
In the platform design model, the software and hardware
platform for each agent are selected and described.
5. IMPLEMENTATION
In the implementation of a prototype of our system we
have followed two steps. The first step consisted in
constructing the user interface, based on an interfaceflow diagram generated from the communication model,
as illustrated in Figure 4. The second step consisted in
implementing the whole process, using ZEUS as our
agent-development platform. It is worth noting that
neither the implementation and test phases are part of
the methodology, since they depend on the employed
platform.
Each interface element activates a process developed by
either an agent or a specific class. We have employed
Java as the programming language for the
implementation of classes and agents.
ZEUS (Nwana et al. 1999) is a system that provides a
design method and tool support for the engineering of
distributed multiagent applications. The provided tools
all encompass the direct-manipulation metaphor and
allow the designer to use drag-and-drop technology to
assemble the application from pre-defined components.
The tool-kit allows the designer to specify models for
different types of agents, for the organizational structure
of agents societies and for negotiation models. A ZEUS
agent has an architecture consisting of five layers: the
definition layer implements the reasoning and learning
capabilities of the agent; the organisation layer
manages and maintains the relationships with other
agents; the coordination layer is responsible for the
inter-agent coordination and negotiation; the
communication layer provides the communication
facilities for the communication with other agents; and finally, the API layer serves as the world interface of the
agents.
THE DEVELOPMENT OF AGENT-BASED E-COMMERCE SYSTEMS
ABSTRACT
This paper illustrates the integration of Guttman’s
Consumer Buying Behaviour model (CBB) into the
MAS-CommonKADS methodology for developing
multiagent systems. We develop each model included in
this methodology for the construction of a recommender
system for buying products from different virtual shops;
and show the integration of the stages of the CBB model
within the MAS-CommonKADS models. We also
describe the implementation of the system using the
agent-building tool ZEUS.
1. INTRODUCTION
Electronic commerce is growing rapidly in the World
Wide Web, and has become a complement to the usual
business activities of corporations and individuals.
Agent technology is considered as the next higher level
abstraction in model-based solutions to e-commerce
applications (Papazoglou 2001). This paper illustrates
the analysis and design of an agent-based recommender
system for buying products from different virtual shops,
using the MAS-CommonKADS methodology (Iglesias
et al. 1998a). The methodology is improved by
incorporating some aspects of the consumer buying
behaviour model of Guttman (Guttman et al. 1998),
when designing e-commerce systems.
The following section describes the consumer buying
behaviour model of Guttman. Section 3 illustrates the
MAS-CommonKADS methodology. Section 4 presents
the analysis and design of the recommender system
using the MAS-CommonKADS methodology,
highlighting how the consumer buying behaviour model
influenced some steps in the methodology. Next, section
5 describes the implementation of a system prototype
using ZEUS (Nwana et al. 1999). Section 6 comparesour work with previous research. Finally, section 7
gathers some concluding remarks and discusses possible
extensions to our work
2. CONSUMER BUYING BEHAVIOUR MODEL
A model of consumer buying behaviour (CBB) usually
comprises the actions and decisions involved in buying
and using goods and services. Typical models focus on
the retail market although they can be related to
business-to-business and consumer-to-consumer
markets as well (Andreasen 1965; Howard and Sheth
1969; Hawkins et al. 1980; Engel and Blackwell 1982).
In this paper we use the model introduced by Guttman
in (Guttman et al. 1998), which extends typical models
with concepts from software agents research in order to
accommodate the growing electronic markets.
The Guttman model proposes six stages that may guide
the consumer buying behaviour in the context of
electronic markets :
• Need Identification. This stage deals with the
identification of consumer’s needs. Within this
stage, the consumer can be stimulated through
product information.
• Product Brokering. This stage is where consumers
determine what to buy. This occurs after a need has
been identified. The decision is made after a critical
evaluation of the retrieved product information. The
result of this stage is an "evoked set" of products.
• Merchant Brokering. This stage combines the
"evoked set" from the previous stage with
merchant-specific information to help determine
who to buy from.
• Negotiation This stage is connected with how to
determine the terms of the transaction.
• Purchase and Delivery. This stage includes the
purchase and delivery of a product.
• Product Service and Evaluation. This stage
involves product service, customer service, and an
evaluation of the satisfactions of the overall buying
experience and decision.
Gutmman shows that these stages occur in several agent
systems for e-commerce, particularly the Product
Brokering, Merchant Brokering and Negotiation stages
(Guttman et al. 1998). For instance, Firefly
(Shardanand and Maes 1995) is a system that helps
consumers find products, as part of the Product
Brokering stage; Jango (Doorenbos et al. 1997) is a
shopping agent that includes price comparisons from
different merchant Web sites, corresponding to the
Merchant Brokering stage; Kasbah (Chavez et al. 1997)
is a multiagent system for consumer-to-consumer
electronic commerce where users create buying and
selling agents to help transact products, automating
much of the Merchant Brokering and Negotiation
stages.
Despite such a wide range of electronic commerce
systems including CBB stages, there is no reference to
the way CBB stages influence the design of multiagent
systems for electronic commerce.
3. THE MAS-CommonKADS METHODOLOGY
MAS-CommonKADS (Iglesias et al. 1998a) is a general
purpose multiagent analysis and design methodology, it
extends the CommonKADS (Schreiber et al. 1999)
design method by combining techniques from objectoriented methodologies and protocol engineering. The
methodology is centred around seven models that cover
the main aspects of the development of multiagent
systems:
• The agent model specifies agents characteristics
such as reasoning capabilities, sensor/effectors,
services, agent groups and hierarchies.
• The task model describes the tasks that the agents
can carry out, such as goals, decomposition,
problem-solving methods, etc.
• The expertise model defines the knowledge needed
by the agents to achieve their goals.
• The organisation model describes the social
organisation of the agent society.
• The coordination model illustrates the conversation
between agents.
• The communication model details the humansoftware agent interactions.
• The design model includes, in addition to the
typical action of the design phase (Pressman 2001),
the design of relevant aspects of the agent network,
selecting the most suitable agent architecture and
the agent development platform.
The application of the methodology consists in
developing the different models. It has been
successfully applied to systems for optimisation of rollmill applications (Iglesias et al. 1998b) and automation
of travel assistants (Arenas and Barrera-Sanabria 2002)
among others.
4. INTEGRATING GUTTMAN CBB MODEL
WITH MAS-CommonKADS
In this section we apply MAS-CommonKADS to the
design of a recommender system for buying books from
different virtual bookshops. The user can look for a
book by entering the author’s name or keywords from
the book title. Then the system shows the options
available from predetermined bookshops, indicating the
cheapest available option. Finally, the user can choose
an option and buy the desired product.
4.1. Conceptualisation
This initial phase aims to carry out an elicitation task in
order to obtain a general description of the problem,
following a user-centred approach based on use cases
(Rambaugh et al. 1999). In this approach, an actor
represents a role played by a person, a piece of
hardware or another system that interacts with our
system. A use case corresponds to a description of the
sequence of actions needed to produce an observable
result useful for an actor.
The CBB model has been useful to identify the actors
and uses cases. The Need Identification stage can be
carried out by including facilities for obtaining client
preferences and informing them about new products in a
proactive way. These facilities are not fully incorporated
in the current version of our system, we have only
included facilities for obtaining basic information about
the actor Client, facilities that are described in the use
case Client Information. In order to confirm the
Product Brokering stage, we include the following
actors: Merchant Agents (one for each participating
bookshop), and a Recommender Agent. A Client makes
a Product Request to the Recommender Agent, who
Requests Merchant Information to each Merchant Agent
in order to obtain the evoked set of products. In the
Merchant Brokering stage the Recommender Agent does
Display Recommendations and Select Best Option.
Traditional retail markets usually do not leave room for
negotiation of price and other aspects of the
transactions. However, a negotiation is central to
markets such as stocks, fine arts, etc. We have not yet
included aspects of the Negotiation stage in our system.
In order to agree with the Purchase and Delivery stage,
we include a Buyer Agent that buys products on behalf
of a client. The Client agent makes a Buy Request, and
the Buyer Agent does Verify Information and performs
the Buy. Finally, the Produce Service and Evaluation
stage can be achieved by means of the evaluation forms
that the Merchant Agent requests to the Client Agent;
this facility will be included in future versions of the
system.
Table 1 summarises the actors and uses cases that we
have identified, as well as the corresponding CBB
stages.
The outcome of the conceptualisation phase is a
description of the different actors and their use cases. As
a way of illustration, Table 2 describes the textual
template of use case Request Merchant Information for
actor Recommender Agent.
Table 2. Textual Template of Use Case Request
Merchant Information for Actor Recommender
Agent
Use Case Request Merchant Information
Summary. The Recommender Agent takes the
information about a book given by the Client and
requests the different Merchant Agents for
additional information (title, price, availability),
giving a list of recommendations.
Actors. Recommender Agent and Merchant Agent
Precondition. Book information and list of
merchant agents.
Postcondition. List of recommendations
Exceptions. No book information; empty list of
merchant agents.
4.2. Analysis
This phase results in the requirement specification of the
system through the development of the first six models
of the methodology. The Agent Model.
The agent model specifies the characteristics of an agent
and plays the role of a reference point for the other
models. An agent is defined as any entity - human or
software - capable of carrying out an activity (Iglesias et
al. 1998a). The identification of agents is based on the
actors and use cases generated in the previous phase of
conceptualisation. The agents we have identified are:
• Client Agent. Human agent who interacts with the
system to obtain information about books available
in the different bookshops.
• Recommender Agent. Software agent that
determines the best shopping option.
• Merchant Agent. Software agent (one for each
participating virtual bookshop) that acts as an interface to the bookshop by offering information
about its available books.
• Buyer Agent. Software agent that buys products
using information provided by the client.
Although we have here an one-to-one association
between actors and agents, this is not always the case.
We can either merge several actors into an agent or
decompose an actor into two o more agents. The
outcome of this model is a set of textual templates (one
for each agent), that shows information such as
description of the agent, parameters, services, etc. For
instance, Table 3 presents the template of the
Recommender Agent.
Table 3. Textual Template of the Recommender Agent
Agent Recommender Agent
Name. Recommender
Type. Software agent
Role. Information provider
Location. Inside the agent society
Description. This agent determines the best
shopping option according to the request made by a
client and the information provided by the
merchants.
Objective. Provides the client with additional
information about the requested book.
Exceptions. Lack of information about a book;
absence of available information in the merchants
about the requested book.
Input Parameters. Book keywords and
information provided by the Merchant Agents.
Output Parameters. List of availability of the
requested book in each bookshop, including the
price and delivering time.
Expertise. This agent must know the profile of the
client who requests a book, in order to apply
heuristics for the recommendation process. It also
stores information about given recommendations so
that the learning process may start.
Communication. Client Agent.
Coordination. Merchant Agent
The Task Model.
The task model describes the tasks that the agent can
carry out. Tasks are decomposed according to a topdown approach. Following (Arenas and BarreraSanabria 2002), we use UML activity diagrams
(Rambaugh et al. 1999) to represent the activity flow of
the tasks. Figure 1 shows the activity diagram for the
Recommender Agent.
Each task is described using a textual template that
includes its name, a short description, input and output
ingredients and required capabilities of the performers.
This documentation is used to support the maintenance
and management of changes in the organisation and
feasibility assessment (Iglesias et al. 1998a). Table 4
shows the textual template of task Determine Best
Shopping Option.
Table 4. Textual Template of Task Determine Best
Shopping Option.
Task Determine Best Shopping Option
Objective. Determine best shopping option from
all the options generated by the virtual bookshops.
Description. Organises the list of products
generated by the Infer Information task from the
cheapest option to the more expensive one. The
cheapest option corresponds to the best shopping
option.
Input Ingredient. List of available products.
Output Ingredient. Best shopping option.
Constraints. None
Exceptions. Absence of products.
The Coordination Model.
The coordination model shows the dynamic
relationships between the agents. It begins with the
identification of the conversation between agents, where
use cases play again an important role. At this level,
every conversation just consists of one single interaction
and the possible answer, which are described by means
of templates as illustrated in Table 5.
Table 5. Textual Template of Conversation Get
Shopping Options
Conversation Get Shopping Options
Type. Getting information.
Objective. Get available books in the virtual
bookshops according to the information provided
by the client.
Agents. Recommender Agent, Merchant Agents
and Client.
Beginner. Recommender Agent.
Service. List of bookshops where the requested
book is available.
Description. The Recommender Agent request for
information about the book to the different
Merchant Agents. The Recommender Agent has a
limited time to get information from other agents.
In case there is a timeout, it sends immediately a
message to the Client notifying it was not possible
to finish the search.
Precondition. Book keywords.
Postcondition. List of bookshops where the book
is available, including price, delivering time and a
short additional information.
Ending condition. Process aborted by the client or
absence of Merchant Agent information.
Next, we model the data exchanged in each interaction
by specifying speech acts and synchronisation, and
collect all this information in the form of sequence
diagrams. In Figure 2, we show the interactions of the
Recommender Agent.
The Communication Model.
The communication model includes interaction between
human agents and other agents. We use templates
similar to those of the coordination model, but taking
into consideration human factors such as facilities for
understanding the recommendation given by the system.
In our system, the communication model includes
interaction between the Client and the Recommender
and the Buyer Agents.
The Expertise Model.
The expertise model describes the reasoning capabilities
of the agents needed to carry out their tasks and achieve
their goals. It consists of two submodels: the
application knowledge and the problem solving
knowledge.
The development of the application knowledge includes
to represent the declarative knowledge of the problem
(domain knowledge), modelled as concepts, properties,
expressions and relationships using the Conceptual
Modelling Language (Schreiber et al. 1999) or graphical
notation of the Object Model of OMT (Milsted 1995);
and to represent the inference steps performed for
solving a task and the order of the inference structures
(inference and task knowledge). In our system, the only
agent that uses knowledge is the Recommender Agent.
We have identified concepts such as books, merchants,
promotions, and properties such as price, book
keywords, etc. These concepts are organised in domain
models that describe particular relationships between
themselves. For instance, we have developed a casual
model of dates of promotions for particular merchants.
The Recommender Agent performs also the generic task
of inferring the best shopping option, including
knowledge about availability of the book in the
bookshops and its price.
The problem solving knowledge includes the
specification of a problem solving method. We use
methods such as assessment that generates schemes for
organising the knowledge (Schreiber et al. 1999). This
method includes the representation of the relevant
knowledge (for instance, special discount for lectures),
the definition of norms of integrity, and its definition in
a representation language.
The Organisation Model.
The study of the organisation is a tool for the
identification of possible impact on the multiagent
system when installed. The organisation model
specifies the structural relationships between the agents
by representing both class agent diagrams and defining
instances of agents. In comparison to the object-oriented
paradigm, the agent-instance information is more
relevant than the class agent diagrams (Iglesias et al.
1998a). Figure 3 shows the class agent diagram of our
system and Table 6 the agent instances.
4.3. Design
In this phase the previous models are used as a basis for
the design model, which consists of the agent network,
the agent design and the platform submodels.
The agent network design model describes the network
facilities (naming services, security, encryption),
knowledge facilities (ontology servers, knowledge
representation translators) and the coordination facilities
(protocol servers, group management facilities, resource
allocation) within the target system. In our prototype
system, we do not use any facilitator agent, so that this
model is not defined.
The agent design model determines the most suitable
architecture for each agent. The method suggests a
generic agent architecture that consists of a usercommunication unit for the agent-user interaction (from
the communication model), an agent communication
unit for the interaction among agents (from the
coordination model) and a deliberation and reaction unit
for the reasoning of the agent (from the agent, expertise
and task models).
In the platform design model, the software and hardware
platform for each agent are selected and described.
5. IMPLEMENTATION
In the implementation of a prototype of our system we
have followed two steps. The first step consisted in
constructing the user interface, based on an interfaceflow diagram generated from the communication model,
as illustrated in Figure 4. The second step consisted in
implementing the whole process, using ZEUS as our
agent-development platform. It is worth noting that
neither the implementation and test phases are part of
the methodology, since they depend on the employed
platform.
Each interface element activates a process developed by
either an agent or a specific class. We have employed
Java as the programming language for the
implementation of classes and agents.
ZEUS (Nwana et al. 1999) is a system that provides a
design method and tool support for the engineering of
distributed multiagent applications. The provided tools
all encompass the direct-manipulation metaphor and
allow the designer to use drag-and-drop technology to
assemble the application from pre-defined components.
The tool-kit allows the designer to specify models for
different types of agents, for the organizational structure
of agents societies and for negotiation models. A ZEUS
agent has an architecture consisting of five layers: the
definition layer implements the reasoning and learning
capabilities of the agent; the organisation layer
manages and maintains the relationships with other
agents; the coordination layer is responsible for the
inter-agent coordination and negotiation; the
communication layer provides the communication
facilities for the communication with other agents; and finally, the API layer serves as the world interface of the
agents.
Sabtu, 12 November 2011
BAB V
SARAN dan KESIMPULAN
Berdasarkan penelitian diatas, maka peneliti menarik kesimpulan atas penelitan yang telah dilakukan dikalangan mahasiswa semoga dapat bermanfaat bagi para pembaca dan semua pihak yang berkepentingan dalam meningkatkan penggunaan provider Indosat.
5.1 SARAN
- Hasil penelitian menyatakan bahwa variabel x2 lebih dominan mempengaruhi penggunaan provider Indosat, maka peneliti menyarankan kepada provider Indosat untuk meningkatkan pelayanannya dengan menambah teknologi pelayanan yang lebih unggul agar para konsumen tidak kecewa akan pelayannan yang meraka dapatkan. Dan agar tingkat penggunaann provider Indosat selalu meningkat.
5.2 KESIMPULAN
- Dari hasil pengamatan yang telah diteliti menggunakan regresi linear berganda Y = 4,182 – 0,463X1 + 0.601 X2 + e dan uji t yang diperoleh untuk variabel harga dan tingkat pelayaan.
- Dari hasil hipotesis sementara yang didapat oleh peneliti, dapat disimpulkan bahwa tingkat pelayanan berpengaruh signifikan yang positif dalam meningkatkan penggunaan provider Indosat.
SARAN dan KESIMPULAN
Berdasarkan penelitian diatas, maka peneliti menarik kesimpulan atas penelitan yang telah dilakukan dikalangan mahasiswa semoga dapat bermanfaat bagi para pembaca dan semua pihak yang berkepentingan dalam meningkatkan penggunaan provider Indosat.
5.1 SARAN
- Hasil penelitian menyatakan bahwa variabel x2 lebih dominan mempengaruhi penggunaan provider Indosat, maka peneliti menyarankan kepada provider Indosat untuk meningkatkan pelayanannya dengan menambah teknologi pelayanan yang lebih unggul agar para konsumen tidak kecewa akan pelayannan yang meraka dapatkan. Dan agar tingkat penggunaann provider Indosat selalu meningkat.
5.2 KESIMPULAN
- Dari hasil pengamatan yang telah diteliti menggunakan regresi linear berganda Y = 4,182 – 0,463X1 + 0.601 X2 + e dan uji t yang diperoleh untuk variabel harga dan tingkat pelayaan.
- Dari hasil hipotesis sementara yang didapat oleh peneliti, dapat disimpulkan bahwa tingkat pelayanan berpengaruh signifikan yang positif dalam meningkatkan penggunaan provider Indosat.
BAB IV
ANALISA dan HASIL PENELITIAN
Pada penelitian digunakan 2 metode untuk menganalisa data primer yang telah diperoleh, yakni metode analisa deskriptif dan metode analisa statistik. Metode analisa deskriptif dalam penelitian ini merupakan uraian atau penjelasan dari hasil pengumpulan data primer yang berupa kuesioner yang telah diperoleh dari responden penelitian. Sedangkan metode analisa statistik , selain untuk menguji uji validitas dan reabilitas dari kuesioner, metode analisa statistik juga digunakan untuk melakukan analisa regresi berganda.
4.1 ANALISA DESKRIPTIF
Instrumen yang digunakan dalam penelitian ini adalah kuesioner. Dari kuesioner tersebut
diperoleh gambaran umum mengenai karateristik responden. Karateristik tersebut meliputi jenis kelamin dan umur.
4.1.1 Karateristik berdasarkan umur
Diketahui bahwa kategori umur yang paling banyak diteliti adalah umur 20 tahun sebanyak 23 orang responden atau 46%. Untuk kategori umur paling sedikit adalah umur 22 tahun sebanyak 27 responden atau sekitar 54%
4.1.2 Karateristik berdasarkan jenis kelamin
Berdasarkan data-data pada kuesioner yang telah disebar oleh peneliti kepada 50 orang responden, diperoleh data mengenai jenis kelamin responden penelitian. Berdasarkan penelitian tersebut dapat diambil kesimpulan bahwa dalam penelitian ini jumalh responden perempuan lebih banyak yakni 31 responden atau 62%, jika dibandingkan dengan responden laki-laki yaitu sebanyak 19 orang yaitu 38%.
4.2 ANALISA STATISTIK
Sebelum kuesioner disebar kepada para responden untuk menjadi sumber data yang baik, perlu diuji layak atau tidak untuk digunakan dalam mengumpulkan informasi bagi penelitian ini. Sehingga data yang diperoleh dari responden dapat digunakan dalam penelitian ini. Oleh karena itu, peneliti mengadakan pra-survei terhadap 30 orang responden untuk menguji kelayakan pernyataan-pernyataan yang akan digunakan dalam penelitian. Kriteria keputusan pengambilan keputusan adalah sebagai berikut:
Untuk mengetahui validitas pada setiap butir pernyataan dalam kuesioner
- Jika r hitung > r tabel maka pernyataan itu valid
- Jika r hitung < r tabel maka pernyataan itu tidak valid
Untuk mengetahui reliabilitas kuesioner:
- Jika r alpha > r tabel maka pernyataan tersebut reliabel
- Jika r alpha < r tabel maka pernyataan tersebut tidak reliabel
Peneliti menggunakan tingkat kesalahan sebesar 5% dan derajat bebas (df) = jumlah butir pertanyaan – 2. Pada uji validitas dan reliabilitas kuesioner dalam penelitian ini diperoleh df = 18-2= 16, maka r reliabel (0,05:16) = 0,468.
4.2.1 Analisis Regresi Linear Berganda
Analisis regresi linear berganda digunakan untuk mengadakan prediksi nilai dari variabel Penggunaan provider Indosat (Y) pada kalangan mahasiswa dengan ikut memperhitungkan nilai variabel pada harga (X1) dan variabel tingkat pelayanan (X2), sehingga dapat diketahui pengaruh positif dan negatif harga dan tingkat pelayanan terhadap penggunaan Provider Indosat di kalangan mahsiswa. Model yang persamaan yang digunakan adalah
Y= a + b1x1 + b2x2 + e
Berdasarkan dari persamaan tersebut diperoleh persamaan regresi linear berganda:
Y = 4,182 – 0,463X1 + 0.601 X2 + e
dari persamaan tersebut dapat digambarkan sebagai berikut:
Konstanta (a) 4,182 menunjukan harga konstanta, dimana jika variabel X1 dan X2 = 0 maka citpenggunaan provider(Y) + 4,182
Dari hasil analisis regresi tersebut diatas terlihat bahwa dari ke dua variabel bebas yang diamati terdapat satu variabel positif dan signifikan terhadap kepuasan konsumen yaitu tingkat pelayanan (X2), sedangkan yang tidak signifikan terhadap kepuasan konsumen terdapat pada harga (X1).
4.3 PEMBAHASAN
Pada hasil Analisis Regresi, koefisien regresi variabel tingkat pelayanan (X2) mempunyai parameter sebesar 0,601, yang ternyata memiliki nilai yang paling tinggi daripada koefisien regresi lainnya. Adapun interpretasi dari parameter tersebut adalah menyatakan bahwa setiap penambahan 1 satuan variabel tingkat pelayanan maka akan meningkatkan kepuasan konsumen sebesar 0,601. Hal ini menunjukkan bahwa variabel tingkat pelayanan secara signifikan berpengaruh terhadap kepuasan. Hal ini berarti faktor tersebut, baik secara parsial maupun kumulatif dengan variabel lain secara signifikan berpengaruh terhadap kepuasan. Implikasinya adalah bahwa untuk meningkatkan penggunaan provider Indosat untuk menjaga keeksistensiannya maka variable tingkat pelayanan dominan perlu diperhatikan oleh perusahaan.
ANALISA dan HASIL PENELITIAN
Pada penelitian digunakan 2 metode untuk menganalisa data primer yang telah diperoleh, yakni metode analisa deskriptif dan metode analisa statistik. Metode analisa deskriptif dalam penelitian ini merupakan uraian atau penjelasan dari hasil pengumpulan data primer yang berupa kuesioner yang telah diperoleh dari responden penelitian. Sedangkan metode analisa statistik , selain untuk menguji uji validitas dan reabilitas dari kuesioner, metode analisa statistik juga digunakan untuk melakukan analisa regresi berganda.
4.1 ANALISA DESKRIPTIF
Instrumen yang digunakan dalam penelitian ini adalah kuesioner. Dari kuesioner tersebut
diperoleh gambaran umum mengenai karateristik responden. Karateristik tersebut meliputi jenis kelamin dan umur.
4.1.1 Karateristik berdasarkan umur
Diketahui bahwa kategori umur yang paling banyak diteliti adalah umur 20 tahun sebanyak 23 orang responden atau 46%. Untuk kategori umur paling sedikit adalah umur 22 tahun sebanyak 27 responden atau sekitar 54%
4.1.2 Karateristik berdasarkan jenis kelamin
Berdasarkan data-data pada kuesioner yang telah disebar oleh peneliti kepada 50 orang responden, diperoleh data mengenai jenis kelamin responden penelitian. Berdasarkan penelitian tersebut dapat diambil kesimpulan bahwa dalam penelitian ini jumalh responden perempuan lebih banyak yakni 31 responden atau 62%, jika dibandingkan dengan responden laki-laki yaitu sebanyak 19 orang yaitu 38%.
4.2 ANALISA STATISTIK
Sebelum kuesioner disebar kepada para responden untuk menjadi sumber data yang baik, perlu diuji layak atau tidak untuk digunakan dalam mengumpulkan informasi bagi penelitian ini. Sehingga data yang diperoleh dari responden dapat digunakan dalam penelitian ini. Oleh karena itu, peneliti mengadakan pra-survei terhadap 30 orang responden untuk menguji kelayakan pernyataan-pernyataan yang akan digunakan dalam penelitian. Kriteria keputusan pengambilan keputusan adalah sebagai berikut:
Untuk mengetahui validitas pada setiap butir pernyataan dalam kuesioner
- Jika r hitung > r tabel maka pernyataan itu valid
- Jika r hitung < r tabel maka pernyataan itu tidak valid
Untuk mengetahui reliabilitas kuesioner:
- Jika r alpha > r tabel maka pernyataan tersebut reliabel
- Jika r alpha < r tabel maka pernyataan tersebut tidak reliabel
Peneliti menggunakan tingkat kesalahan sebesar 5% dan derajat bebas (df) = jumlah butir pertanyaan – 2. Pada uji validitas dan reliabilitas kuesioner dalam penelitian ini diperoleh df = 18-2= 16, maka r reliabel (0,05:16) = 0,468.
4.2.1 Analisis Regresi Linear Berganda
Analisis regresi linear berganda digunakan untuk mengadakan prediksi nilai dari variabel Penggunaan provider Indosat (Y) pada kalangan mahasiswa dengan ikut memperhitungkan nilai variabel pada harga (X1) dan variabel tingkat pelayanan (X2), sehingga dapat diketahui pengaruh positif dan negatif harga dan tingkat pelayanan terhadap penggunaan Provider Indosat di kalangan mahsiswa. Model yang persamaan yang digunakan adalah
Y= a + b1x1 + b2x2 + e
Berdasarkan dari persamaan tersebut diperoleh persamaan regresi linear berganda:
Y = 4,182 – 0,463X1 + 0.601 X2 + e
dari persamaan tersebut dapat digambarkan sebagai berikut:
Konstanta (a) 4,182 menunjukan harga konstanta, dimana jika variabel X1 dan X2 = 0 maka citpenggunaan provider(Y) + 4,182
Dari hasil analisis regresi tersebut diatas terlihat bahwa dari ke dua variabel bebas yang diamati terdapat satu variabel positif dan signifikan terhadap kepuasan konsumen yaitu tingkat pelayanan (X2), sedangkan yang tidak signifikan terhadap kepuasan konsumen terdapat pada harga (X1).
4.3 PEMBAHASAN
Pada hasil Analisis Regresi, koefisien regresi variabel tingkat pelayanan (X2) mempunyai parameter sebesar 0,601, yang ternyata memiliki nilai yang paling tinggi daripada koefisien regresi lainnya. Adapun interpretasi dari parameter tersebut adalah menyatakan bahwa setiap penambahan 1 satuan variabel tingkat pelayanan maka akan meningkatkan kepuasan konsumen sebesar 0,601. Hal ini menunjukkan bahwa variabel tingkat pelayanan secara signifikan berpengaruh terhadap kepuasan. Hal ini berarti faktor tersebut, baik secara parsial maupun kumulatif dengan variabel lain secara signifikan berpengaruh terhadap kepuasan. Implikasinya adalah bahwa untuk meningkatkan penggunaan provider Indosat untuk menjaga keeksistensiannya maka variable tingkat pelayanan dominan perlu diperhatikan oleh perusahaan.
Langganan:
Komentar (Atom)
