Type-2 Fuzzy Web Shopping Agents Menglei Tang and Yanqing Zhang Georgia State University Gang Zhang Tianjin University
Introduction(1) n n Fuzzy logic was first invented as “representation schema and calculates for uncertain or vague notion. ” Type-2 fuzzy set: grade of membership is a fuzzy set in [0, 1]. Type-1 fuzzy set : membership is a crisp number in [0, 1]. A type-2 fuzzy logic system includes 5 components: fuzzifier, fuzzy rule, fuzzy inference, type reducer and defuzzifier. Its antecendent and consequent set are all type-2 and it has type reduction to map a type-2 set into a type-1 set.
Introduction(2) n n n an online shopping system “hotstore. com” is created and a type-2 fuzzy logic system is designed. the basic architecture of the fuzzy online shopping data mining system. the area-weighted fuzzy reasoning method used for the fuzzy online shopping data mining system. the real implementation. the analysis of the fuzzy online shopping data mining method. concludes this work.
Fuzzy Online Shopping Agents(1) n 2 1 3 Customer Fuzzy Rule Based System Web Service(IIS) 3 Database(SQL) n Database(SQL): defines shopping system data, which include merchandise information, customer information (login record, customer detail), and order information. Fuzzy rule based system: provides an integrated environment for fuzzy data mining and decision support
Fuzzy Online Shopping Agents(2) Procedure: n n n 1. Customer visitor the shopping system, such as login, browser, buy merchandise etc through IE and the page forward the request to web server. Web server fetches the information to IE. 2. Web server gets, changes and inserts the date from SQL server database. 3. Web server calls fuzzy data mining system to calculate merchandise rating by historical data stored in database.
Area –Weighted Type-2 Fuzzy Reasoning(1) Number Rating
Area –Weighted Type-2 Fuzzy Reasoning(2) Type-2 Fuzzy Logic system 1. Fuzzifier. It maps crisp input into a fuzzy set. The crisp input of type-2 fuzzy logic system is merchandise total bought number by the customer (called number for simplicity). 2. Fuzzy rule. If-then rules whose antecedent and consequent set are type-2. 3. Inference Engine. It combines rule and gives a mapping from type-2 input to type-2 output. 4. Type reducer. It maps a type-2 input into a type-1 output. n 5. Defuzzifier. It maps a type-1 set into a crisp set which is merchandise rating.
Area –Weighted Type-2 Fuzzy Reasoning(3): Fuzzifier n Type-2 primary number 0 to 50 membership function n 0 to 100 Fuzzy Set u(x) Low 1 -x/50 Middle Low x/50 if 0<x<50 2 -x/50 if 50<x<100 1. 0 50 to 150 Middle x/50 -1 if 50<x<100 3 -x/50 if 100<x<150 ui L ML M MH H 100 to 200 Middle Large X/50 -2 if 100<x<150 4 -x/50 if 150<x<200 x>150 0 50 x 100 150 200 Large X/50 -3 if 150<x<200 1 if x>200
Area –Weighted Type-2 Fuzzy Reasoning(4): Fuzzifier n The secondary membership function If number=x, it’s primary membership is ui. The secondary membership function is 1. 0 0 Xm x where
Area –Weighted Type-2 Fuzzy Reasoning(5): Inference Engine n n n Inference includes a set of rules. Rule 1: If number is low, then rating is low. Rule 2: If number is middle low, then rating is middle low. Rule 3: If number is middle, then rating is middle. Rule 4: If number is middle high, then rating is middle high. Rule 5: If number is high, then rating is high.
Area –Weighted Type-2 Fuzzy Reasoning(6): Type Reducer n n n Type reduction produces a type-1 fuzzy set from type-2 fuzzy set. The method used here is Centroid Type Reduction. Let assume that input primary member ship function is ui, its secondary membership function is for . For firing strength=, it type reducer value is
Area –Weighted Type-2 Fuzzy Reasoning(6): Defuzzification converts the partial fuzzy conclusions generated by type reducer to a crisp value. n Centroid dufuzzifier method can be expressed by: if Partial Fuzzy Conclusion is Small, if Partial Fuzzy Conclusion is Middle, if Partial Fuzzy Conclusion is Middle Large, if Partial Fuzzy Conclusion is Large. n
Example Here gives an example when merchandise is sold for 60 items, that is number=60. Step 1: Fuzzy Matching: , n Step 2: Inference Engine Fuzzy Rule 2: Firing strength of Middle Low ; Fuzzy Rule 3: Firing strength of Middle. Step 3: Type Reducer Step 4: Defuzzifier
Implementation n n Home page: a welcome page for the customer. Customer Pages: include add_an_account page and login page. Merchandise Page: lists merchandise detail information, including merchandise ID, merchandise images, prices, detailed information. Cart Page: lists merchandise in shopping cart. Checkout Page: lists all the products the customer selects.
Conclusion Designing an online shopping system “hotstore. com” n Mining web data and merchandise rating using type-2 fuzzy logic system. n Benefits of fuzzy Web shopping data algorithm : (1) software can implement it easily (2) Business logic can be built into the fuzzy rules (3) automate the decision making process for e. Commerce applications. n
Скачать презентацию Type-2 Fuzzy Web Shopping Agents Menglei Tang and
96a24fbc78de1b7fe45c80013a304cfc.ppt