8/21/2003INDIN'2003, Workshop on Soft Computing...1 BISC Decision Support System Masoud Nikravesh BISC The Berkeley Initiative in Soft Computing Electrical Engineering and Computer Sciences Department The basic ideas underlying soft computing in its current incarnation have links to many earlier influences, among them Prof. Zadeh ’ s 1965 paper on fuzzy sets; the 1973 paper on the analysis of complex systems and decision processes; and the 1979 report (1981 paper) on possibility theory and soft data analysis. The principal constituents of soft computing (SC) are fuzzy logic (FL), neural network theory (NN) and probabilistic reasoning (PR), with the latter subsuming belief networks, evolutionary computing including DNA computing, chaos theory and parts of learning theory. Fuzzy Set: 1965 … Fuzzy Logic: 1973 … Soft Decision: 1981 … BISC: 1990 … Human-Machine Perception: …
8/21/2003INDIN'2003, Workshop on Soft Computing...2Outline BISC Decision Support System System components Applications Web-Based BISC DSS Multi-Criteria Querying Model: EC- based optimization
8/21/2003INDIN'2003, Workshop on Soft Computing...3 BISC Decision Support System Objectives: Develop soft-computing-based techniques for decision analysis Tools to assist decision-makers in assessing the consequences of decision made in an environment of imprecision, uncertainty, and partial truth and providing a systematic risk analysis; Tools to assist decision-makers answer “What if Questions”, examine numerous alternatives very quickly and find the value of the inputs to achieve a desired level of output; Tools to be used with human interaction and feedback to achieve a capability to learn and adapt through time;
8/21/2003INDIN'2003, Workshop on Soft Computing...4 BISC DSS: Components and Structure Model Representation Including Linguistic Formulation Functional Requirements Constraints Goals and Objectives Linguistic Variables Requirement Input From Decision Makers Model Management Query Aggregation Ranking Fitness Evaluation Evolutionary Kernel Genetic Algorithm, Genetic Programming, and DNA Model and Data Visualization Data Management Selection Cross Over Mutation Experts Knowledge
8/21/2003INDIN'2003, Workshop on Soft Computing...5 BISC DSS: Process of Expert System User User Interface Dialog Function Knowledge Base Editor Inference Engine Recommendation, Advice, and Explanation Knowledge Refinement Data IF … THEN Rule Expert Knowledge users ask for advice or provide preferences inferences & conclusion advises the user and explains the logic expertise is transferred and it is stored Knowledge Base
8/21/2003INDIN'2003, Workshop on Soft Computing...6 BISC DSS: Data & Knowledge Management Data Sources and Warehouse (databases) Knowledge Representation, Data Visualization and Visual Interactive Decision Making Knowledge Discovery and Data Mining Knowledge Generation Knowledge Bases Organization Expert Knowledge
8/21/2003INDIN'2003, Workshop on Soft Computing...7Applications Financestock prices and characteristics, credit scoring, credit card ranking Military battlefield simulation and decision making Medicinediagnosis Marketingstore and product display electronic shopping Internet provide knowledge and advice to large number of users Education university admissions Bankingfraud detection
8/21/2003INDIN'2003, Workshop on Soft Computing...8 Case : Profitable Customers A computer system that uses customer data that allow the company to recognize good and bad customer by the cost of doing business with them and the profits they return keep the good customers improve the bad customers or decide to drop them identify customers who spend money identify customers who are profitable compare the complex mix of marketing and servicing costs to access to new customers
8/21/2003INDIN'2003, Workshop on Soft Computing...9 Case: Fraud Detection An Intelligent Computer system that can learn the user’s behavior through in mining customer databases and predicting customer behaviours (normal and irregularities) to be used to uncover, reduce or prevent fraud in credit cards stocks financial markets telecommunication insurance
8/21/2003INDIN'2003, Workshop on Soft Computing...10 Web-Based BISC Decision Support System BISC The Berkeley Initiative in Soft Computing Electrical Engineering and Computer Science Department Gamil Serag-Eldin, Masoud Nikravesh
8/21/2003INDIN'2003, Workshop on Soft Computing...11 Web-based DSS: objectives Existing search system models using crisp logic and queries objects need to match exactly the decision criteria which results in rigid systems with imprecise and subjective process and results Objective: develop a multi-criteria fuzzy querying model
8/21/2003INDIN'2003, Workshop on Soft Computing...12 Web-based DSS : Design Conceptual level Resembling natural human behavior - allowing approximation –objects do not need to match exactly the decision criteria Implementation level Designed in a generic form to: –accommodate more diverse applications –to be delivered as stand-alone software to academia and businesses.
8/21/2003INDIN'2003, Workshop on Soft Computing...13 Web-based DSS Components Fuzzy Search Engine (FSE), Application Templates (AT), User Interface (UI), Database (DB), Computational Intelligence (CI).
8/21/2003INDIN'2003, Workshop on Soft Computing...14 Web-based DSS: general framework UI DB Fuzzy Search Engine (FSE) Application Template (AT) Computational Intelligence (CI) Aggregators Similarity measures Membership functions
8/21/2003INDIN'2003, Workshop on Soft Computing...15 User interface & Application template A specific HTML interface and template for each application we developed. UI DB Fuzzy Search Engine (FSE) Input mapping Control unit
8/21/2003INDIN'2003, Workshop on Soft Computing...16 Database (DB) This module handles all queries or user’s profile creations from the User Interface and the Fuzzy Engine respectively. DB Fuzzy Search Engine (FSE) UIUI DB Manager User Profile Query
8/21/2003INDIN'2003, Workshop on Soft Computing...17Applications Credit Scoring Date Matching University Admissions Diagnosis
8/21/2003INDIN'2003, Workshop on Soft Computing...18 Multi-Aggregator Fuzzy Decision Tree: EC-based optimization BISC The Berkeley Initiative in Soft Computing Electrical Engineering and Computer Science Department Souad Souafi-Bensafi, Masoud Nikravesh
8/21/2003INDIN'2003, Workshop on Soft Computing...19 Multi-Criteria Querying (1) Multi-Attribute Query Database x 12 x 11 x 1k x 22 x 21 x 2k x N2 x N1 x Nk q2q2 q1q1 qkqk Query Similarity calculation Scores S j2 S j1 S jN Query Answering Ranking based (criteria: number top answers) Selection based (criteria: threshold) Scoring model
8/21/2003INDIN'2003, Workshop on Soft Computing...20 Multi-Criteria Querying (2) Scoring model: Calculation of similarity between data and query: similarity measures for crisp or fuzzy data are calculated for each attribute and combined using aggregation operators to provide a global score User preferences Represented in the scoring model by the parameters: similarity measures, aggregation operators and corresponding parameters (weights, combination strategies) Decision making process Data Multi-attribute query
8/21/2003INDIN'2003, Workshop on Soft Computing...21 First-order aggregation model (1) Model decription S(x 1, x 2, …, x k ) x2x2 x1x1 xkxk … q1q1 q2q2 qkqk … Aggregator w1w1 w2w2 wkwk … similarities measures AggregationScore queryweights
8/21/2003INDIN'2003, Workshop on Soft Computing...22 First-order aggregation model (2) User’s preferences representation limited to weights associated with attributes Optimization process : find the optimal weights Using GA. GA-based learning module Optimal weights Specific fitness function First-order aggregation model Problem specification w1w1 w2w2 wkwk … Model parameters learning using GA
8/21/2003INDIN'2003, Workshop on Soft Computing...23 Advanced multi-aggregation model Model description Parameters - aggregators, - weights and - tree structure. Aggregation tree Aggregators Attributes Model parameters learning using GP GP-based learning module Optimal multi- aggregation model Specific fitness function Aggregators, Attributes Specific DNA encoding Problem specification
8/21/2003INDIN'2003, Workshop on Soft Computing...24 Fitness calculation (1) Input fuzzy data Similarity calculation ( 1 (Q) | | n (Q) ) Query Score calculation For each attribute Aggregation Tree For each data row x i Score ( x i ) Score Ranking For each Aggregation Tree ( 1 (x i ) | | n (x i ) )
8/21/2003INDIN'2003, Workshop on Soft Computing...25 Fitness calculation (2) Min YES Max NO good answers = Min YES Max NO Fitness function combines : distance to maximize Min YES Max NO bad answers good answers Overlap <= 0Separation > 0 bad answers Tree size to minimize Score Ranking