S. Stolfo et. al, The ALEXSYS mortgage pool allocation expert system: A case study of speeding up rule-based programs. Columbia University Department of Computer Science and Center for Advanced Technology,  Home/Search   Document Details and Download   Summary   Related Articles   Check

....consists of hard rule systems that investigate an extensive amount of data. The complexity of these problems requires robustness from both the rule inference engine and the data retrieval utility. Representative applications include network security monitors and real time decision control systems [85,94,95] (See Appendix) The goal of this research is to facilitate hard active database applications by abstracting language limitations that inhibits their development. The contributions of this dissertation are threefold. First, the previous generations of active database languages were designed for ....

....to module as the result of a jump, or a jump to subroutine, executed in the action of a rule. Consequently, only one module at a time is sensitive to the current 1. Hard active database technology been available at the time of the program s development, the technology would have been exploited [85] 31 state of the program. Thus, not all possible alternatives are evaluated at each cycle. The correctness of the program requires the rule developer to consider both data driven and procedural methods of programing. The motivations behind VenusDB are most similar to those that produced the ....

[Article contains additional citation context not shown here]

S. Stolfo et. al, The ALEXSYS mortgage pool allocation expert system: A case study of speeding up rule-based programs. Columbia University Department of Computer Science and Center for Advanced Technology,

.... active database applications that we have developed ranging across point of sale, medical patient, network security monitors, real time decision control systems, and process control monitors can be classified into a subclass of hard rule systems called monotonic log monitoring (MLM) applications [5,11,20,24,25,26] . MLMs process real time data logged to a database. The primary reason a DBMS is chosen is to exploit the database s query and data durability services as a platform for decision support. A fundamental property of the MLM logs is that they are inserted to, but never updated nor deleted. It is ....

S. Stolfo et. al, The ALEXSYS mortgage pool allocation expert system: A case study of speeding up rule-based programs. Columbia University Department of Computer Science and Center for Advanced Technology, 1990.

....applications. The precise definition and implementation of VenusDB is the result of examining application needs in the database arena. The ALEXSYS mortgage pool allocation program, an expert system now widely deployed in the financial securties industry is our most thoroughly exploited application [14,13]. The migration of these applications from batch oriented, stand alone programs to reactive multidatabase platforms demonstrated to us a clear need to distinguish some database updates as triggerable events and others as not. The rule language then had to be extended to include an explicit trigger ....

....a condition and executing an action are significantly more costly (main memory speed vs. disk access speed) reducing the number of rule evaluations is imperative. Application Requirements Our analysis of application requirements is based largely on our experience with the ALEXSYS program [14]. That program is from the domain of finance. It takes items from inventory (collections of individual mortgages with similar characteristics called pools) and matches them with orders (orders to buy mortgages called contracts) according to a variety of government regulations. Pools can be split ....

Stolfo, S. and et.al. "The ALEXSYS Mortgage Pool Allocation Expert System: A Case Study of Speeding Up Rule-based Programs." Columbia University Department of Computer Sciences and Center for Advanced Technology, 1990.

.... a Real World Rule based System This section investigates the issues involved in parallelizing Alexsys, a rule based system developed at Columbia University in conjunction with Citicorp to handle the problem of fulfilling contracts to trade pools of mortgages in such a way as to maximize profit [30] . Alexsys represents an expert system which has been developed for use in the real world with actual data and differs markedly 23 from the small benchmark programs discussed previously in this work in terms of the complexity of the rules and the size of the working memory database. However, it ....

....profit and minimizing the number of surplus unallocatable pools and unassigned contracts. There are many subtle aspects of the allocation process that are not germane to the discussion at hand; for a more complete description of the Alexsys system and its potential for parallel speedup, see [28, 30] . In Alexsys, the control structure of the process by which pools are assigned to contracts is divided into three major parts: initialization, pool allocation, and report generation. During the initialization phase, the most profitable unallocated contract is selected and assigned the ....

[Article contains additional citation context not shown here]

Salvatore J. Stolfo, Leland Woodbury, Jason Glazier, and Philip Chan. The ALEXSYS mortgage pool allocation expert system: A case study of speeding up rule-based programs. In AI and Business Workshop, AAAI-90, 1990.

....used. One example is intersection spatial joins [Brinkhoff et al. 1994] where a possible set of spatial objects are first identified and then a more complex geometric filter is applied to determine which objects satisfy the spatial join predicate. Another example is the strategy used in ALEXSYS [Stolfo et al. 1990], an expert system for allocating mortgage pools, where pools that can be successfully allocated can be found close to one another if an initial order is given to the data. Determining if two or more pools can be allocated to a contract is determined by the rules in the expert system. A third ....

S. Stolfo, L. Woodbury, J. Glazier, and P. Chan. The ALEXSYS Mortgage Pool Allocation Expert System: A Case Study of Speeding up Rule-based Systems. In AI and Business Workshop, AAAI-90, 1990.

....any serializable execution schedule is as good as any other serializable schedule. For rule processing, serializability is not an adequate constraint on rule executions, since very often one is interested in expressing heuristic preferences between rule instances that may not have any shared data [80]. Thus data locking or other methods for ensuring serializability in a concurrent execution environment is not appropriate for a rule language like PARULEL, which must provide the means for expressing applications that are epistemologi 57 cally and heuristically adequate for a large class of ....

S. J. Stolfo, L. Woodbury, J. Glazier, and P. Chan. The ALEXSYS mortgage pool allocation expert system: A case study of speeding up rule-based systems. AI and Business Workshop, AAAI-90, 1990.

....for settlements to be arithmetically correct. The Public Securities Administration has published many pages of detailed constraints limiting how pools are assigned to buy orders; not the least of which is that orders need not be filled exactly but within 2. 5 of the value of the order[23]. ALEXSYS exploits 8 different heuristic method regulatory combinations to define and fulfill buy orders. For the purposes of presentation, we describe an abstracted and simplified version of ALEXSYS that we have also implemented that we called Sawmill. Logs arrive in batches at a sawmill for ....

....elaborated into final executable code, much like top down designs. Stepwise refined programs give rise to many forms of the formal methods of program development; any refinement can be proven to be correctness preserving, that is, one refinement correctly specifies the next refinement [23]. Stepwise refinement was used to implement REALESYS, the Venus implementation of ALEXSYS, in this case study. 5 The OPS5 Implementation of the Mortgage Pool Allocation Problem, ALEXSYS After deciphering the organization of the secret messages in ALEXSYS, it is possible to expose the modular ....

S.J. Stolfo et al., "The ALEXSYS Mortgage Pool Allocation Expert System: A Case Study of Speeding UpRule-based Programs, " Columbia University Department of Computer Science and Center for Advanced Technology, New York, NY, July 1990.

....miranker cs.utexas.edu 1. Introduction In the 1980s and early 1990s, numerous expert system applications were fielded in industry. Many of these applications inferenced over data stored in databases. One large application of this type that we have studied is the ALEXSYS mortgage pool allocator [1,2]. Recently, research on the integration of rules and databases has centered on active databases, which offer tighter coupling between the database and the rule system. One research prototype is our system Venus [3,4,5] an outgrowth of the Datex project [6] This paper reports on our experience ....

....a technique where common cases are handled by the new active database application and anomalous cases are handled by the existing expert database application may be an appropriate first step. 2. Background 2. 1 ALEXSYS The ALEXSYS mortgage pool allocation program is from the domain of finance[1]. It takes items from inventory (in this case, collections of individual mortgages with similar characteristics called pools) and matches them with orders (in this case, orders to buy mortgages called contracts) according to a variety of government regulations. Pools can be split into smaller ....

Stolfo, S., et.al. "The ALEXSYS Mortgage Pool Allocation Expert System: A Case Study of Speeding Up Rule-based Programs." Columbia University Department of Computer Science and Center for Advanced Technology, New York, NY. July 1990. Available from http://www.cs.columbia.edu/~sal.

....buy orders. We have implemented a solution to the mortgage pool allocation problem in Venus, coined REALESYS. We then developed quantitative software metrics and compared their values against the same statistics developed for the original implementation in the OPS5 rule language, coined ALEXSYS [STO90, WAR96]. The OPS5 version has also been recoded in C and widely licensed by financial institutions. It would clearly be valuable to measure the C version as well, but that version is proprietary. 5.2.1 REALESYS REALESYS was designed top down, using stepwise refinement. The REALESYS solution used the ....

Stolfo, S. et.al. "The ALEXSYS Mortgage Pool Allocation Expert System: A Case Study of Speeding Up Rule-based Programs." Columbia University Department of Computer Sciences and Center for Advanced Technology, 1990.

....ALEXSYS program is concise (a total of 51 rules and 19 functions) and is the end product of nearly a full year of knowledge engineering and prototyping. Most of this time was devoted to investigating several techniques used to speed the running time of the OPS5 program from days to seconds [6]. The human allocators strategy is straightforward and was replicated as closely as possible by ALEXSYS. The humans use a greedy algorithm (appropriately named for this application) to search the space of partial allocations where, at each branch point, an irrevocable decision is made to choose ....

S. J. Stolfo, L. Woodbury, J. Glazier, and P. Chan, "The ALEXSYS mortgage pool allocation expert system: A case study of speeding up rule-based programs," in Proceedings of the AI in Business Workshop, AAAI, 1990.

....the problem. The result is a rule based program that solves the problem better than human experts. However, at present, the rule based system is slower than desired for very large data sets and has been reimplemented in C to achieve acceptable performance. The system, which we call ALEXSYS [29], serves as our testbed for various studies on parallelism. The mortgage pool allocation problem is faced by financial companies that trade mortgagebacked securities. An individual security is a collection of homeowner mortgages, commonly referred to as a pool. A contract is an agreement between ....

....and has been used and approved by banking professionals. Second, it is a program moderate in size, which makes it easier to analyze. Third, after various optimization techniques have been applied to ALEXSYS, it is still undesirably slow when large amounts of data are presented to the system [29]. Lastly, the mortgage pool allocation problem possesses characteristics that challenge some of the traditional parallelization techniques, which forced us to invent ways to tackle these newly surfaced problems. The organization of ALEXSYS, its inherent sequentialities, and attempts to parallelize ....

[Article contains additional citation context not shown here]

Stolfo, S. J., Woodbury, L., Glazier, J., and Chan, P. The ALEXSYS mortgage pool allocation expert system: A case study of speeding up rule-based systems. AI and Business Workshop, AAAI-90, 1990.

No context found.

S. J. Stolfo, L. Woodbury, J. Glazier, and P. Chan, The ALEXSYS mortgage pool allocation expert system: A case study of speeding up rule-based systems, AI and Business Workshop, AAAI-90, (1990).

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC