The Trust In Ecommerce

Published Date: 02 Nov 2017

Introduction

Trust is related to uncertainty in e-commerce environment. Every new technology has few pros and cons; internet is also part of this new scientific innovation that has revolutionized our life. Business houses are largely relying on this new technology. Internet has become integral part of our lives these days. But there exists darker side to this innovation; the use of e-commerce has brought about new set of electronic threats and risks [74]. These risks include misuse of personal data (e.g. credit card number), web spoofing and deliberate misinformation. This has led to building of distrust [19, 33] with the e-commerce framework. A certified server authenticates the server not the web content information on the server. An authentic server only shows web uses that it is truly the server it claims it is. Similarly the certificate shown by the website does not guarantee the product information it displays, it is only concerned about transaction security in real time mode. The metadata present in Trust Modeling [73] represent the trustworthiness of a web page and Trust Attributes.

In this chapter we first focus on trust as defined by various authors and later develop a Fuzzy Regression model [15, 23] of various Trust Attributes. We will be highlighting the work of Yang, Brown and Lewis [121] in identifying the Trust Attributes. We also develop our model TrFRA (Trust Based Fuzzy Regression Analysis).

3.1 Defining Trust

Trust, as we know it, is a very important aspect of human life. We use it every day in one way or another. For instance, we go to work and trust airplanes not to crash our houses meanwhile, we trust our bank to give us the money we claim and some even trust their computers to work every time they switch them on. Trust is central to all transactions, where our own actions are dependent on the actions of others. Thus, excluding instances where trust in someone has no influence on our decisions. Trust can be strong or weak depending on the environment. Morton Deutsch defines trust as follows [60]:

If an individual is confronted with an ambiguous path, a path that can lead to an event perceived to be beneficial (Va+) or to an event perceived to be harmful (Va-);

He perceives that the occurrence of Va+ or Va- is contingent on the behavior of another person; and

He perceives that strength of Va- to be greater than that strength of Va+.

In this definition, Deutsch describes an event with a beneficial outcome as Va+, while a harmful result is entitled as Va-. His view of a trust relationship is such that any event is linked with other events by either conducive or detrimental paths. To reach another event, one of the available paths has to be taken. This definition is based on psychology but outlines one of the most important requirements to establish trust. If the perceived benefit were greater than the perceived harmfulness then the significance of trust in the choice would not be that big. In other words, a trust relation requires a harmful path with more significance than the beneficial one. An employee for example, would not choose to hack a database and switch identity with another employee if the others salary was less than his own. A more concrete and mathematical definition is given by Diego Gambetta [24]:

"Trust (or, symmetrically, distrust) is a particular level of the subjective probability with which an agent assesses that another agent or group of agents will perform a particular action, both before he can monitor such action (or independently or his capacity ever be able to monitor it) and in a context in which it affects his own action.

When we say we trust someone or that someone is trustworthy, we implicitly mean that the probability that he will perform an action that is beneficial or at least not detrimental to us is high enough fur us to consider engaging in some form of cooperation with him.

Correspondingly when we say that someone is untrustworthy, we imply that that probability is low enough for us to refrain from doing so."

Mcknight and Chervany [81] as inspiration Josang et al. [59, 60] define trust as:

"Trust is the extent to which one party is willing to depend on somebody, or something, in a given situation with a feeling of relative security, even though negative consequences are possible."

This definition may be vague but it includes all the relevant components of a trust interaction. Beside the above definitions and restraints, trust is treated differently in some other sectors. Sociology for example deals with responsibilities and moral in this context. These concepts, although strongly related to the notion of trust, are hard to measure or to include in a conceptual form. The ability to measure trust is the main prerequisite to create a concept of a trust-based security environment in a network.

By saying "we trust someone" or that "he is trustworthy", we without reservation mean that probability of his being performing a task that is constructive or at least not destructive to us is high enough for us to consider him associating with us in some form of cooperation.

Similarly by saying that "someone is untrustworthy", we mean that that probability of him being associating with us is extremely low.

In a network-mediated virtual world, where interaction is represented by the exchange of messages between digital entities, trust is a fundamental challenge; identity is suspect, the exchange medium is suspect, and there is often no history of interaction or expectation of future interaction between particular entities. Nevertheless, trust remains a requirement for maintenance of cooperative social groups and online economic activity.

The communication between server and client are not secure unless it is providing a safe and secure transaction. To reduce the risk we have to deal with development of trustworthiness of the web services, which finally means increasing the complexity of the website.

3.2 Trust in E-Commerce

Trust is considered imperative and a main concern within an e-business because the customers do not interact with the business or the staff face-to-face. In e-commerce one must create trustworthy relationships, attract and maintain a customer base within e-businesses due to the absence of the physical product, where the customer and seller are physically separated. This can result in an insecure environment where trust is of high importance. Companies need to create relationships which result in trust relationships where initial sales will be generated, which can lead to customer loyalty.

Trust is often equated with security concerns in online transactions. A secure transmission implies that the two parties in a transaction have been properly authenticated and that the information exchanged via the network remains unaltered. However, there are three main ways in which confidential information can be obtained [14]:

Information copied during transmission

Information accessed during storage

Information obtained from an authorized party

Whom can you trust? Apparently, 95% of all security incidents are caused by insider attacks [10]. This means that secure systems that have been properly set up are still at risk from people who have legitimate access to the system.

A major problem facing the full deployment of business-to-consumer (B2C) e-commerce is the development of trust on the side of the consumer. People develop trust in a business through their own previous experience with that business or through reports about that business from trusted third parties or other consumers. While a great deal of effort has been spent on privacy and secure transactions, resulting in seals of approvals and trustmarks, not much has been done in the that will provide assurance to the consumer that the e-commerce site with which they are dealing is "legitimate" in such things as delivery, return policies, etc. A problem with seals and trustmarks is that they are an "all-or-nothing" approach and give a single value (the seal). A more flexible system is required that rates e-commerce sites along multiple dimensions and allows the consumer to determine which dimension is of importance at that moment in time.

There are many reasons for the slower growth in B2C than in B2B e-commerce, but one reason has to do with the much slower development in e-commerce supporting services (such as security and payment services) on which B2C e-commerce relies. These supporting services are important in creating the "legitimacy" conditions required for trust to develop on the part of the consumer. Some important "legitimacy" conditions [62] that allow trust to develop include:

the sellers are who they claim to be

the seller has right of sale over the item in question

the transaction and payment mechanisms are available, legal and secure

information about the buyer is not redistributed to other organizations or used for other purposes than for which it was intended

the item sold corresponds to its description and is suitable for its intended purpose

the purchased item can and will be delivered to the buyer

the buyers are who they claim to be

the buyer has the resources to purchase the item

All but the last two items are issues of "trust" from the perspective of the consumer, while the last two items are issues of trust from the perspective of the seller. In this research, we have concentrated on the issue of trust from the perspective of the consumer.

3.3 Current Approaches for Evaluating Trust

Current work on trust issues is highly segmented, with individual groups concerned more about their own particular fields of inquiry. Few of these approaches are discussed below.

3.3.1 Self-Regulation

On-line trade communities can self-regulate, and thereby encourage the growth of trust among consumers, by participating in reputation-based schemes that provide a seal (trustmark), once a site has satisfied minimum trust criteria. The seal is withdrawn if there are any violations.

The W3 Consortium’s Platform for Privacy Preferences Project (P3P) attempts to provide a framework for informed online interactions and provides a way for a website to encode its data-collection and data-use practices in a standardized, machine-readable XML format.

The two most popular trust label programs are TRUSTe (www.truste.org) and BBBOnLine (www.bbbonline.org). TRUSTe is a nonprofit organization whose mission is to build user trust in the Internet by promoting the principle of disclosure. In the TRUSTe system, privacy is the main trust criterion. Following an online organization’s request, TRUSTe audits that organization’s website. If a website adheres to established privacy principles, i.e. meets the core criteria, and is willing to comply with oversight and consumer resolution procedures, then the TRUSTe seal will be awarded to the website.

Similarly, the BBBOnLine Privacy program offers a ‘seal’ to websites that post online privacy policies and meet the principles of the Better Business Bureau (disclosure, choice and security). It also monitors compliance and applies specific sanctions for non-compliance.

Trust label programs require vigilance in their monitoring to ensure that privacy standards are upheld. However recent surveys have showed that people do not seem to understand privacy seal programs.

A negative reputation system [87] has been proposed in which information on untrustworthy traders is publicly distributed. In other words, reputation serves both as a source of information and as a potential source of sanctions. These systems collect, distribute, and aggregate feedback about buyers’ and sellers’ past behavior. Though few of these people know each other, the system helps them to decide who to trust, thereby encouraging trustworthy behavior, and deterring dishonest participants.

From a social perspective, Friedman et al [8] believe ‘people trust people, not technology’. They also suggest online trust can be cultivated through 10 trust-related characteristics of online interaction, such as

reliability and security of the technology;

knowing what people online tend to do;

misleading language and images;

disagreement about what counts as harm;

informed consent;

anonymity;

accountability;

saliency of cues in the online environment;

insurance and performance history and

reputation

They point out that we are vulnerable to trust violations in two ways: loss of money and loss of privacy.

3.3.2 Presentation of Website

Several empirical works on the importance of online trust factors for keeping customer loyalty [2, 29, 33, 36, 68, 88, 108, 111] have identified a number of customer-interface design and trust factors, including assurances, references, certifications, privacy provisions, consumer protection, and security policies.

Head et al, have an understanding of the way trust is built through humanized website design [61, 82]. They believe there is a connection between human elements in user interface design. Online trust and User Confidence can be built through a humanized website featuring several factors: brand, assurances, fulfillment and website design.

3.3.3 Agent-Based User Technology

Agent technology researchers [6, 22, 27] have tried to develop ‘smart and intelligent’ agents that could be combined with other technology such as ‘virtual reality’ to facilitate trust based on ‘promises being made, enabled and fulfilled’ to build a trusted customer relationship between customers and providers. This research considers the cornerstone for a successful and lasting relationship with the customer is trust, as it could determine the customer’s future behavior and loyalty towards the business.

Although the work reveals some interesting ways of implementing web trust through agent technology, Web users may have difficulty accepting this kind of external manifestation of trust agent, which does not address the issue of why technologies should be trusted. In addition, replacing Web-based interfacing with an agent-based interface may not be feasible for a considerable time because of inertia between the implementation of a particular approach and acceptance by general Web users. However, over time this mismatch could be overcome as a result of increased knowledge by Web users.

3.3.4 Mathematical Approaches

A number of trust models including Josang 1999 and Abdul-Rahman & Hailes mainly address the problem of an entity’s identity by using cryptographic mechanisms for propagating trust measures that take place within the information security community.

Josang’s ‘beliefs’ model is based on subjective logic, which is an extension of standard logic and to a certain degree, probability theory. The model may be a suitable technique for assigning trust values in the face of uncertainty. The proposed back-propagation [80] method tries to automatically generate a metadata description, and make it easier to classify Web information by ‘fuzzifying’ the metadata attributes.

3.3.5 Public Key Infrastructure and Security

A number of trust models use public key encryption to provide a Web authentication framework, and attempt to achieve the maximum of trust with the minimum of risks. These trust models include the X.509 standard Public Key Infrastructure (PKI) trust model, the Pretty Good Privacy (PGP) trust model, the Simple Public Key Infrastructure (SPKI) and the Simple Distributed Secure Infrastructure (SDSI). However it is important to note that PKI trust models could not achieve 100 per cent trust. As Gutmann stated, ‘CA certificates can exhibit numerous vulnerabilities. Because all CAs are assigned the same level of trust, the entire system is only as secure as the least secure CA.

Nonetheless, there is a need for a range of PKI trust models, from the formal hierarchy used in X.509 to the Web of trust used by PGP and a much-simplified data structure used by SDSI & SPKI. These PKI trust models provide different structures of trust and take different approaches to establishing a trust relationship. Their certification services also provide a much-needed variety of levels of trust among business parties on the Web.

Many researchers propose various approaches to address different trust issues in the eCommerce environment, including a better human-oriented front-end design, improved technologies for online transaction, and public user protection policies. However, the trustworthiness of webcontent cannot be addressed by online security technologies alone.

Chapter – 4

Trust v/s complexity of e-commerce sites

Introduction

India today is facing with various kinds of threat to e-commerce systems. The problem arises when we increase the security of the e-commerce website, the complexity at the user level also increases, which in turn affects the volume of sale. While traditional marketing does not involve any type of complexity since the consumer deals directly with the supplier. Since internet marketing does not involve any face to face direct interaction so a visual interface is essential. There are various types of online buying behavior models like Bettman (1979) and Booms (1981) in which the focus was on personal characteristics viz. Culture, Social Group and Physiological Behavior. Lewis and Lewis (1997) have classified web users in five categories:

Directed information-seekers: It includes those users who are looking for product information only and normally not planning to buy online.

Undirected information-seekers: Such users commonly referred as 'surfers', generally ends up on a particular website accidentally by browsing and following hyperlinks. Users of this group tend to be inexperienced users and also likely to click banners of the website.

Directed buyers: These are actual buyers, who are online to purchase specific products. Such users would visit location that compares product features and prices.

Bargain hunters: These users love to hunt the offers available from sales promotions such as free samples or contests.

Entertainment seekers: Such users interact with web for seeking fun through entering contests such as quiz’s, puzzles or interactive games.

Under all the above categories the main focus is the trust of web users which will finally lead to purchase.

The communication between server and client are not secure unless it is providing a safe and secure transaction. To reduce the risk we have to deal with development of trustworthiness of the web services, which finally means increasing the complexity of the website.

Fig. 4.1. Trust/ Complexity Matrix.

From the above figure we can conclude that there has to be some situation in which a trade off between Trust Level and Complexity of the transaction has to be maintained. This trade off can be achieved by the help of development of Fuzzy Rule base, but simple Fuzzy Rule base will not be sufficient for this purpose, so we extend this problem and solve it using Evolutionary Multi-objective Optimism [42, 79, 94].

4.1 Trust Issues in Ecommerce

Web trust issues include User Confidence, privacy, online security, and authenticity of business partners, service providers and products. Trustworthiness of the webcontents could be viewed as a foundation of Web trust. Web trust can be described as a counterweight to elements of uncertainty. All Web communities have overwhelmingly agreed that ‘eCommerce is a matter of trust’.

The Web is becoming increasingly important in providing the infrastructure for electronic commerce. The Web offers an unprecedented business opportunity for small- and medium-size businesses, and more and more of them are taking advantage of the low cost of establishing a Web-based business.

Electronic Commerce (eCommerce) has changed the conventional service relationship between business parties (i.e. consumers and service providers) and consequently, the traditional way of assessing trust in a business relationship. For example, the traditional way of verifying legitimate business identities was based on the physical identity, such as the shopfront of a business provider. On the Web, business providers and consumers may no longer be able to be identified by traditional means (i.e. their physical appearance); instead, their identity is manifested in their websites, email addresses or by some electronic means (e.g. an electronic token, commonly a public key or digital ID).

These changes have brought a range of potential risks to Internet users, including fraud, misuse of personal data (e.g. credit card numbers), deliberate misinformation, Web spoofing (i.e. mimicking legitimate businesses in order to unlawfully obtain credit card numbers), eavesdropping, identity theft, repudiation, unlawful webcontent modification, masquerading and insecure transmission. These perceived threats are partly the result of a lack of rules, online service standards, codes of conduct (or protocols) and the ability to police them, creating ‘uncertainty’ in the eCommerce environment. Some of the elements of this uncertainty are easily identified, while some are hidden or embedded in the eCommerce environment (e.g. misinformation hidden in the webcontent). This uncertainty affects consumer confidence in the online ecommerce.

Trust requirements can be based on various consumer expectations and user-confidence improvement factors. A national survey of Internet users for consumers by Webwatch identified nine deciding factors that Web users rely on when visiting a website:

ease of navigation;

being able to trust the information on a website;

being able to easily identify the sources of information on a website;

knowing the website is updated frequently with new information;

being able to find out the important facts about a website;

knowing who owns the website;

knowing what businesses and organizations financially support the website;

the presence of seals of approval from other groups; and

the presence of awards and certificates from other groups.

The above surveys provide empirical evidence of Web users’ perspectives on online trust and trustworthiness requirements, for the benefit of both web users and providers, i.e. for service providers, a better user-interface design; for Web users, a basic assessment of the trustworthiness of a website.

However following services are important in creating the legitimacy conditions required for trust to develop on the part of the consumer. Some important legitimacy conditions that allow trust to develop include:

the sellers are who they claim to be

the seller has right of sale over the item in question

the transaction and payment mechanisms are available, legal and secure

information about the buyer is not redistributed to other organizations or used for other purposes than for which it was intended

the item sold corresponds to its description and is suitable for its intended purpose

the purchased item can and will be delivered to the buyer

the buyers are who they claim to be

the buyer has the resources to purchase the item

All but the last two items are issues of "trust" from the perspective of the consumer, while the last two items are issues of trust from the perspective of the seller. In this research, we have concentrated on the issue of trust from the perspective of the consumer.

4.2 Existing Web Based Trust Models

Many trust-related projects involve researchers, practitioners, industries and governments. The World Wide Web continues its efforts in every possible Web-related field to advance Web technology. The following projects have explored various ways to address Web trust.

The REFEREE Project (REFEREE: Rule-controlled Environment for Evaluation of Rules, and Everything Else) [114] by W3C working groups. It provides a general mechanism for expressing and assessing trust management policies. The mechanism may be used to solve problems related to trust management that exist in the World Wide Web by building a trust infrastructure i.e. protocols, policy, languages, execution environment and metadata format for all Web applications requiring trust. REFEREE checking user policies in response to host application’s request for action. Policies are generally regarded as programs in REFEREE. For any given request, REFEREE calls upon the appropriate user policy and interpreter module and returns the host application a reply to the question of whether or not the request fulfill with the policy.

The DSig project [101]: The DSig project uses digitally signed labels to make authenticate able assertions about standalone documents or about manifests of aggregate objects.

PICS [86] was developed by the W3 Consortium [W3C96] to address the problems of protecting children from pornography on the Internet without violating the right to freedom of speech.

The PolicyMaker [13] which is based on PICS and was originally designed to address trust management problems in network services that process signed requests for action and use public-key cryptography. It specifies what a public key is authorized to do [12].

KeyNote [11], the successor to Policy Maker, was developed by AT&T Research laboratories to improve on the weaknesses of PolicyMaker with two additional design goals: standardization and ease of integration.

Recreational Software Advisory Council: This model is used by parents and teachers for filtering the content of information on the Web. In 1999 it was ‘folded into’ a new organization, the Internet Content Rating Association (ICRA). The original aims of RSAC, to guard children from potentially harmful content while safeguarding free speech on the Internet, continue to provide the cornerstone for ICRA’s work. The RSACi system (RSAC on the Internet) has been incorporated into Netscape Navigator and Microsoft’s Internet Explorer, the latter since the release of version 3.0 in February 1996.

Much research has focused on the area of consumer confidence and behaviors. From the Web consumer’s viewpoint, likely concerns could be related to privacy, security, authenticity of the service providers, and trustworthiness of published information on websites.

4.3 Our Proposed Model

Genetic Algorithms [25, 84, 119] have been frequently used to model a solution for conflicting goals. Let Trust (T) be a measure of security which the customer will be provided and Inverse of Complexity (C) be the user comfort level. Applying the Fuzzy Rule base we can get

Maximize Trust (T) (4.1)

But it leads to compromise in the complexity (C) of fuzzy rule based systems [56, 57, 109, 119]. According to consumers survey most of consumers in India considers Trust and Ease of Use (Lower level of Complexity) at the same time. The above problem can be formulated as

Maximize Trust (T) subject to

Inverse complexity (C) (4.2)

where complexity (C) is the measure of fuzzy rule system.

We can develop a single objective function to the above solution given as:

Maximize Æ’(Trust (T), Inverse of Complexity (C)) (4.3)

We can also use weights in order to determine the exact function for e-commerce site.

Maximize (w1) Trust (T) +

(w2) Inverse of Complexity (C) (4.4)

We proceed with development of more refined stages in which we can focus on various stages of membership functions. Consider a simple single output function y = f(x) an application of Takagi-Sugeno method [77, 109, 110, 119] we can write it as:

Rule Ri : if x is Ai then y=ai+bjx, i=1,2,…N

Rule Rk : if x is Ak then y=ak+bkx, k=1,2,…N

:

:

:

Rule Rz : if x is Az then y=az+bzx, z=1,2,…N (4.5)

This output value is given as:

(4.6)

where y(x) is the estimated output value for the input value x and µAi (x) is the membership value of the antecedent fuzzy set Ai.

From the input-output data we can derive the relationship between Trust and Complexity of the e-commerce site considering three Takagi-Sugeno Rules.

We develop a heuristic rule [102, 109] denoted by three lines A, B and C as the subsequent of the linear function with fuzzy sets A1, A2 and A3. Each of the Fuzzy Rule can be represented in triangular Fuzzy Sets.

Rule R1: If TRUST is SMALL and COMPLEXITY is HIGH Then User’s Ease of Use is MEDIUM.

Rule R2: If TRUST is LARGE and COMPLEXITY is MEDIUM Then Users Ease of Use is HIGH.

Rule R3: If TRUST is SMALL and COMPLEXITY is SMALL Then Users Ease of Use is HIGH.

Fig. 4.2. Three Takagi-Sugeno Rules

Based on the above rules we try to develop a plot between Complexity and Trust and develop our interpretable solution [40, 65, 78, 118, 120] between the two entities.

Fig.4.3. Input Output Data using Fuzzy Data Set

Possibly we can also merge the above set of rules to achieve more refined results, but a relationship generated by optimization rules gives some gridlines in the area of relationship between the two entities.

4.4 Conclusion

It is very difficult to interpret the exact relationship between the two entities. Different Fuzzy rule are being applied in order to determine the appropriate interpretability. The method that we have used is the application of Fuzzy Optimization Theory [54, 55, 64, 76] to find the probable relation-ship between Complexity and Trust. The future extension would be to use Evolutionary Algorithm [57, 66] in finding out the best possible trade-off between the two entities.

***

Chapter – 5

WEB PERSONILIZATION AND RECOMMENDATION

5.0 Introduction

Electronic Commerce is fast emerging as most popular method of purchasing, let it be a small pen drive or bulky LED TV. Recent survey [51] has estimated that around 3-5% of Indians have transacted or are well versed with working of online shopping websites. The strategy which is being followed until now related to the various policy initiatives like:

Consumer Proportion: This model is being propagated by the government based on certain guidelines for the protection of consumers.

Legality: It deals with formal recognition of electronic signatures; In India digital signatures are necessary for e-Tendering.

Security: Central Government has issued its policy relating to cryptography techniques to ensure secure electronic commerce in third party transfer.

In order to deal with security and web personalization [17] issues we develop two basic classification methods: Naïve Bayes and K-nearest neighbor. We start this chapter with the introduction to Web Personalization and Recommendation System. Later in this chapter highlights the classification methodologies using Bayesian Rule for Indian e-commerce websites. It deals with generating cluster of users having fraudulent intentions. Secondly, it also focuses on Bayesian Ontology Requirement for efficient Possibilistic Outcomes.

5.1 Web Personalization & Recommendation System

Web personalization, consisted of activities such as providing customized information, changing the webpage layouts and adapting the contents tailored to the user’s need, has become an essential part of a website to enhance its compatibility and attractiveness. The recommendation system [103] or interactive decision aids [43] can be considered as one form of personalization to facilitate in helping the users making purchase decisions.

Main differences between the traditional brick-and-mortar stores and e-commerce websites are the infinite shelf-space on the Web. Unlike the traditional stores which have limited storage, the E-commerce websites provide the consumers a wide variety of options, alternatives and product information. The diversity of product choices and the abundance of messages on an e-commerce site have led to the problem of overloading. To overcome this problem demand of web personalization and real-time adaptation catering to the user’s need has arise. Reason being shopping experience can be overwhelming especially when there is no assistance available in deciding what products to purchase. In addition, the effort and time spent on searching aimlessly may lead to poor quality of decision and dissatisfaction of the consumers [16]. Therefore, to find the ideal products in mind effectively and efficiently, online customers not only look for the suggestions from their peers, and editorial picks [107] but also heavily count on the real-time recommendation systems featured on the e-commerce websites [43, 106].

Recent Web technological advances help online companies to acquire individual customer’s information in real time. Based on this information, they construct detailed profiles and provide personalized services. Thus e-shops now have the opportunity to improve their performance by addressing individual user preferences and needs, increasing satisfaction, promoting loyalty, and establishing one-to-one relationships.

A recommendation system is a system or application that helps the user to select a suitable item or finding relevant information among a set of candidates using a knowledge-base that can either be hand coded by experts or learned from behaviors of the users. Typically, a recommendation system performs three of functions:

Information Collection: The recommendation system collects all the usable information for the prediction task including the users’ attributes, behaviors, or the content of the resources the user accesses.

Learning: It applies a learning algorithm to filter and exploit the users’ features from the collected information.

Prediction: It implies the kind of resources the user may prefer are then made either directly based on the dataset collected in the information collection phase (memory-based predictions) or with a model learned from it (model-based predictions).

Recommender systems are used by E-commerce sites to suggest products to their customers and to provide consumers with information to help them decide which products to purchase. The products can be recommended based on the top overall sellers on a site, on the demographics of the consumer, or on an analysis of the past buying behavior of the consumer as a prediction for future buying behavior. The forms of recommendation include suggesting products to the consumer, providing personalized product information, summarizing community opinion, and providing community critiques. Broadly, these recommendation techniques are part of personalization on a site because they help the site adapt itself to each customer.

Recommender systems are similar to, but also different from, marketing systems and supply-chain decision-support systems. Marketing systems support the marketer in making decisions about how to market products to consumers, usually by grouping the consumers according to marketing segments and grouping the products in categories that can be aligned with the marketing segments. Marketing campaigns can then be run to encourage consumers in different segments to purchase products from categories selected by the marketer. By contrast, recommender systems directly interact with consumers, helping them find products they will like to purchase. Recommender systems include processes that are conducted largely by hand, such as manually creating cross-sell lists and actions that are performed largely by computer, such as collaborative filtering.

Recommender systems enhance E-commerce sales in three ways:

Converting Browsers into Buyers: Visitors to a Web site often look over the site without purchasing anything. Recommender systems can help consumers find products they wish to purchase.

Increasing Cross-sell: Recommender systems improve cross-sell by suggesting additional products for the customer to purchase. If the recommendations are good, the average order size should increase. For instance, a site might recommend additional products in the checkout process, based on those products already in the shopping cart.

Building Loyalty: In a world where a site’s competitors are only a click or two away, gaining consumer loyalties is an essential business strategy. Recommender systems improve loyalty by creating a value-added relationship between the site and the customer. Sites invest in learning about their customers, use recommender systems to operationalize that learning, and present custom interfaces that match consumer needs. Consumers repay these sites by returning to the ones that best match their needs. The more a customer uses the recommendation system – teaching it what he wants – the more loyal he is to the site.

5.1.1 Types of Recommendation System

Types of recommendation system can be classified based upon techniques used for recommending

5.1.1.1 Content-Based Filtering

The Content-Based Filtering (CBF) makes recommendation based on the correlation between difference resources. In content-based recommendation systems, resources are described as a vector of attributes. The system then learns profile of the users interests based upon the features presented in the objects that user has rated. When making a prediction on the customers’ preferences, the system analyzes the relationship between the products rated by the users and other products by calculating the similarity between their attribute vectors. The type of user profile derived by a content-based recommender depends on the learning method employed. Decision trees, neural nets, and vector-based representations have all been used.

A central problem in content-based recommendation systems is the need to identify a sufficiently large set of key attributes. When the set is too small, there is insufficient information to learn the customer profile. Therefore, content-based recommendation systems cannot be used for new customers who purchased only once, potential customers who visit the web site but have not made any purchase, and customers who want to buy a product that is not frequently purchased.

5.1.1.2 Collaborative Filtering

The collaborative filtering (CF) is widely implemented and most mature of the information filtering technologies. Collaborative recommendation systems aggregate ratings or recommendations of objects, recognize commonalities between users on the basis of their ratings, and generate new recommendations based on inter-user comparisons. A typical user profile in a collaborative system consists of vector of items as well as their ratings, continuously modified as the user interacts with the system over time.

Collaborative filtering algorithms are classified into two classes: memory-based and model-based. Memory-based algorithms operate over the entire user database to make predictions. The most common memory-based models are based on the notion of nearest neighbors, using a variety of distance measures. Model-based systems are based on a compact model inferred from the data, which have used a variety of learning techniques including neural networks, latent semantic indexing and Bayesian networks.

The greatest strength of collaborative techniques is that they are completely independent of any machine-readable representation of the objects being recommended. In addition, they work well for complex objects such as music and movies, where variations in taste are responsible for much of the variation in preferences.

The main difference between collaborative and content-based filtering systems is that the collaborative systems track past actions of a group of customers to make a recommendation for individual members of the group. Using this approach, customers may now be able to receive recommendations for products that are dissimilar in content to those they have previously rated, as long as other like-minded customers showed their interests in these products.

The collaborative filtering identifies customers whose interests are similar to those of a given customer, and recommends products of the given customer have liked. However, as most existing CF algorithms strongly depend on the user’s ratings (on items to make recommendations, their performances decay dramatically when the user rates few items in the database, which is called the new user or cold start problem in the CF research.

5.1.1.3 Knowledge-Based Recommendation

The knowledge-based recommendation attempts to propose objects based upon inferences about users needs and liking. Knowledge-based approaches are distinguished in that they have functional knowledge: "they have knowledge about how a particular item meets a particular user need", and can therefore reason about the relationship between a need and a possible recommendation. The user profile can be any knowledge structure that supports this inference. In the simplest case, as in Google, it may simply be the query that the user has formulated. In others, it may be a more detailed representation of the user’s needs.

5.1.1.4 Utility-Based Recommendation

The utility-based recommendation does not attempt to build long-term generalizations about their users, but rather base their advices on an evaluation of the match between the user’s needs and the set of options available. The utility-based recommendation makes suggestions by working out the utility of each object to the user.

The benefit of utility-based recommendation is that it can factor non-product attributes, such as vendor reliability and product availability, into the utility computation, making it possible, for example, to trade off price against delivery schedule for a user who has an immediate need.

5.1.1.5 Demographic Recommendation

The demographic recommendation systems aims to classify the user based upon personal attributes and then make suggestions based on demographic classes. The users’ responses are matched against a library of manually assembled user stereotypes.

The representation of demographic information in a user model can vary greatly. Demographic techniques form "people-to-people" correlations like collaborative ones, but use different data. The benefit of a demographic approach is that it may not require a history of user ratings of the type needed by collaborative and content-based techniques.

5.3.1 Advantages & Disadvantages of Naïve Bayes Classifier

An advantage of the Naive Bayes Classifier is that it only requires a small amount of training data to estimate the parameters (means and variances of the variables) necessary for classification. Because independent variables are assumed, only the variances of the variables for each class need to be determined and not the entire covariance matrix. The logic of using Naïve Bayes Classification Technique [115] is to attain computational efficiency and good performance.

5.3.2 Fuzzy Information Classification and Retrieval Model

The above section deals with a classification technique [105] by which we can categorize the customer visiting our site based on their transaction history. The characteristics of Fuzzy systems that give them better performance for specific applications are:

Fuzzy systems are suitable for uncertain or approximate reasoning, especially for a system with mathematical model that is difficult to derive.

Fuzzy logic allows decision making with estimated values under incomplete or uncertain information.

In this section we have highlighted the problem which our customer face while selecting the best possible combinations of product, the problem is because of the uncertainty in Semantic Web Taxonomies [39]. Consider Indiatimes shopping portal shown in figure 5.1.

Fig. 5.1: Indiatimes Shopping Portal

If a buyer wants a laptop in the range of Rs.25000 < x < Rs.35000, and with features F = {f1, f2, f3} in brands B = {b1, b2}, then he must be shown the best possibilistic outcome of the above query.

The above problem looks very simple but it is not so, there exists an uncertainty in the query, what if, if there is no laptop with all the features of ‘F’ present in Brand ‘B’. Here comes a probabilistic method to overcome such situation.

In our method, degrees of subsumption will be covered by Bayesian Network based Ontology’s [28]. The Venn diagram is shown in figure 5.2.

Fig. 5.2: Venn Diagram Illustrating Electronic Items with Laptops as one of their Categories & their Overlap

Our method enables the representation of overlap between a selected concept and every other is referred taxonomy. The Price Range-I represent the prices at the start of the price band while Price Range-II represents the higher side of the price band.

The overlap is logic term expressed as

(5.4)

The overlap region represents the value 0 for disjoint concepts and 1, if the referred concept is subsumed by the selected one. This overlap value can be used in information retrieval tasks. The match with the query is generalized by the probabilistic sense and the hit list can be sorted into the order of relevance accordingly.

If ‘F’ and ‘B’ are sets; then ‘F’ must be in one of the following relationships to ‘B’.

‘F’ is a subset of ‘B’ i.e. .

‘F’ partially overlaps ‘B’ i.e.

‘F’ is disjoint from ‘B’ i.e.

Based on these relations we develop a simple transformation algorithm. The algorithm processes the overlap graph G in a Breadth First manner starting from root concept defined as ‘CON’. Each processed concept ‘CON’ is written as the part of Solid Path Structure (SPS).

if F subsumes B then

O := 1

else

C = Fs ∩ Bs

if C =  then

O := 0

else

∑ m(C)

O :=

end

end

Fig. 5.3: Computing the Overlap

The overlap values ‘O’ for a elected concept ‘F’ and a referred concept ‘B’.

If F is the selected concept and B is referred one, then the overlap value 0 can be interpreted as the conditional probability

(5.5)

where S(F) and S(B) are taken is and interpreted as a probability space, and the elements of the sets are not interpreted as elementary outcomes of some random phenomenon.

The implementation stages of the probabilistic search starts with the Input of Ontology Rule which are refined in "Refinement Stage". It is than passed to the "Quantifier" which develops a set of Association Rules. It is then fed to the further preprocessing by the "Naïve Bayesian Transformation" module which finally generates the best possible overlapping outcome as shown in figure 5.4.

Fig. 5.4: Implementation Framework.

5.4 Conclusion

The model in both the cases uses interactive query refinement mechanism to help to find the most appropriate query terms. The Ontology is organized according to narrower term relations. We have developed an algorithm in which taxonomies can be constructed without virtually any knowledge of Probability and Bayesian network. The future extension could be to expand it using Fuzzy Regression [122] with Bayesian Network.

***

Chapter - 6

CONCLUSION

We would like to summarize our work by highlighting various features of Web Personalization and Recommendation Model for Trust in Ecommerce Website from an Indian Perspective.

We have studied that potential for growth of e-commerce in India is enormous. Some findings like amount of interest that is there for online travel industry is not seen in case of other services. People prefer non tangible goods i.e. services over tangible products. Professional e-commerce websites are doing excellent job but still there are some factors that are inhibiting users from purchasing online. After survey and interview we concluded that for users to adopt e-commerce, it is vital that the benefits of using this commercial medium (e.g. convenience, saving in time and transaction costs) significantly overshadow potential risks. Indeed, the user's freedom to select appropriate vendors needs to be correlated with greater concerns regarding financial risk, privacy and trust. This can be accounted for by the fact that private users are directly involved in the commercial exchange; they are using their own equipment, giving sensitive information about themselves as individuals, and spending their own money. There are still various factors to be looked upon to cater to needs of the consumer who is the driving force for e-commerce. TRUST, CONVENIENCE, SECURITY are prime factors without which we will not be able to attain our goal.

In our quest for building Trust in "TrFRA: A Trust Based Fuzzy Regression Analysis", the major thrust of the model is to find out trust building factors in e-commerce websites. It also focuses on fuzzy relationship between Trust and website related factors. TrFRA is a relationship model in terms of fuzzy output and regression line. The output represents Trust Factor and its correlation with two factors WOC (Web Object Content) and WOCA (Web Object Certification Authority) also the estimated value for the output variable. TRUST is derived by the difference between the observed and the estimated value is assumed to be due to the ambiguity inherently present in the system. The output Trust for a specified input is assumed to be a range of possible values i.e. output can take on any of the possible values. The advantage with Fuzzy Regression is the range of Possibilistic values is much more as compared to the Normal Regression Model.

Second phase of our work focuses on Trust on a website which leads to increasing level of securities, this might irritant user enough to migrate on other websites. In "Trust Vs Complexity of E-commerce Sites", we have highlighted how we need to tradeoff between Trust and Complexity so as not to drive away users from the website. It is very difficult to interpret the exact relationship between the two entities. Different Fuzzy rule are being applied in order to determine the appropriate interpretability. We have applied Fuzzy Optimization Theory to find the probable relation-ship between Complexity and Trust. The future extension of this problem uses Evolutionary Algorithm to find out the best possible trade-off between the two entities.

Our work on "Web Personalization & Recommendation" highlights how to identify fraud users by using Classification Methodologies of Bayesian Rules and generating cluster of users having fraudulent intentions. The model in both the cases uses interactive query refinement mechanism to find the most appropriate query terms. The Ontology is organized according to narrower term relations. We have developed an algorithm in which taxonomies can be constructed without virtually any knowledge of Probability and Bayesian network. The future extension could be to expand it using Fuzzy Regression with Bayesian Network.

Most companies involved in e-commerce use various trust-promoting methods to persuade consumers their websites are safe. Although these methods are common, many consumers still hesitate to carry out transaction online. In this thesis a sincere effort has been made to understand why and how the extent to which trust-promoting methods can be used to change consumers’ attitudes toward online purchasing. Our findings will certainly help shopping websites analyze their existing trust-promoting methods and improve them by adding persuasive elements to enhance the trust and their sales.

***