![]() |
|||
![]()
|
![]() |
![]() |
![]() |
6. MENSWEAR CASE: VARIABLESThe decision variables for the configuration of the menswear case are listed in Figure 2. In this study, we consider five product types (suit, shirt, slacks, necktie, and jacket). Each product attribute corresponds to a variable. For the product-related variables, when the values that meet the customer's requirement for a certain criteria are identified, the Product Matcher should dynamically retrieve the items that meet the requirement from the product database, as illustrated in Figure 2. For each access of a given customer, the customer information is initialized by retrieving the data from the customer database as revealed in the example in Figure 3. For an advanced system, we may wish to store the history of purchase and learn to guess what each particular customer will need, but this is beyond the scope of the current version.
FIGURE 2 Variables in menswear case. 7. MENSWEAR CASE: CONSTRAINTSThe constraints can be classified by structural characteristics. Note that the constraints have an importance indicator to set their priorities relative to each other. 7.1. IMPORTANCE OF CONSTRAINTSThe importance of each constraint is marked in the importance slot ranging between 0 and 1. The measurement of "1" implies "must," and "0" implies "not necessary." The following constraint implies that the compatibility between customer's face color and shirt color must be met. {{ Constraint-C9 Importance: 1 Associated-variables: customer.face-color shirt.color Compatible-value-set: (pale (pink light-blue)) }}
FIGURE 3 Representation of a product.
FIGURE 4 Representation of a customer. 7.2. TYPES OF CONSTRAINTSFrom a structural perspective, the constraints in UNIK-SES can be classified into four types: value compatibility, value incompatibility, algebraic inequality, and functional assignment. 7.2.1. Value Compatibility Constraints The value compatibility constraints are the what-matches-what relationships between products. The following example shows a constraint, which means the "wide-spread" collar of a shirt goes well with the "double-breasted" suit. {{ Constraint-C1 Importance: 1 Associated-variables: shirt.collar suit.breasted Compatible-value-set: (wide-spread double-breasted) }} Another type of value compatibility constraint would be general tendencies for matching of customers and products based on the salesman's experience. For instance, the occupation may influence the style of a shirt as the following example shows, but the importance level of this kind of constraint should be weaker. {{ Constraint-C6 Importance: 0.5 Associated-variables: customer.occupation shirt.collar buying-plan.purchasing-purpose Compatible-value-set: (salary-man regular office-work-use) }} 7.2.2. Value Incompatibility Constraints On the other hand, there are also many constraints that prohibit the what-does-not-match-what incompatible relationships in the configuration domain. The following example shows an incompatibility constraint: a heavy man does not look well when he wears a suit with horizontal stripes. {{ Constraint-I1 Importance: 1 Associated-variables: customer.body-type suit.pattern Incompatible-value-set: (heavy horizontal-stripe) }} 7.2.3. Functional Assignment Constraints Functional assignment constraints represent the assignment relationships between independent variables and a dependent variable. For instance, the total amount of a sale is computed by summing five price components. The reverse directional inference cannot be applied. {{ Constraint-F1 Dependent-variable: buying-plan.total-amount Independent-variable: suit.price shirt.price slacks.price neck-tie.price jacket.price Relation: buying-plan.total-amount = suit.price + shirt.price + slacks.price + neck-tie.price + jacket.price }} 7.2.4. Algebraic Inequality Constraints The budget constraint of a customer, for instance, can be represented using algebraic inequality constraints. {{ Constraint-A1 Importance: 0.3 Associated-variables: buying-plan.total-amount buying-plan.total-amount-upper-limit Relation: buying-plan.total-amount >= buying-plan.total-amount-upper-limit}} 8. RULES IN MENSWEAR CASEThe purpose of rules used in Salesman Expert System can be classified into customer-product rules and sales strategic rules. 8.1. CUSTOMER-PRODUCT RULESWhile a constraint handles the what-does-not-match-what kinds of relationships without any directional nature, a rule has a directionality such as "if a jacket is going to have patterns, the slacks should not have any." The reverse relationship in general does not hold. {{ Rule-1 IF (jacket.pattern IS checkered) THEN (slacks.pattern IS no-pattern) Importance: 1 }} The role of the importance slot in the rules is the same as in the constraints. 8.2. SALES STRATEGIC RULES FOR VARIABLE AND VALUE ORDERINGSales strategies are represented as variable and value ordering rules that play the role of meta-rules in CRSP reasoning. 8.2.1. Variable Ordering Rules An experienced salesman may ask a student customer the affordable price range first, but may ask a 40-year-old lady customer whether she wants a one-piece or two-piece first. Being able to ask quesions in the right sequence is a skill that is essential for business success. In CRSP terms, the sequencing of questions is the variable ordering. An example is : {{ Variable-ordering-rule-1 IF (income-level IS high)) THEN (seed.variable IS brand) Importance: 0.5 }}
|
![]() |
|
Use of this site is subject certain Terms & Conditions. Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details. |