MARS: A mergers and acquisitions reasoning system

Mergers and acquisitions (M&A) are currently revolutionizing the structure of corporate U.S.A. and annually involve deals totalling billions of dollars. Consequently, it is an area of intense activity and interest within the financial community. The process of planning an M&A is enormously complex and involves sophisticated reasoning and planning, by several parties such as the raider, the target company, investment banks, etc. Computer based tools are often invaluable for planning several stages of an M&A, such as generating forecasted cash flows. Current computer aids for M&A however do not provide adequate support for many essential features such as real time planning, reasoning under uncertainty, nonmonotonic inference, case-based reasoning, etc. MARS is a prototype M&A reasoning tool developed at General Electric Corporate R&D that attempts to provide such features in an integrated environment. MARS both simulates and provides advice regarding the complex reasoning and planning involved in an M&A deal. In doing so, it provides an excellent test bed architecture for the testing, development and integration of several ideas from artificial intelligence. MARS is implemented in COMMON LISP using RUM [15] on top of KEE [18]. RUM, a development environment for reasoning under uncertainty is based on Bonissone's theory of plausible reasoning [2–4] and was also developed at General Electric Corporate R&D.     

A classification and regression technique to handle heterogeneous and imperfect information

Imperfect information inevitably appears in real situations for a variety of reasons. Although efforts have been made to incorporate imperfect data into learning and inference methods, there are many limitations as to the type of data, uncertainty and imprecision that can be handled. In this paper, we propose a classification and regression technique to handle imperfect information. We incorporate the handling of imperfect information into both the learning phase, by building the model that represents the situation under examination, and the inference phase, by using such a model. The model obtained is global and is described by a Gaussian mixture. To show the efficiency of the proposed technique, we perform a comparative study with a broad baseline of techniques available in literature tested with several data sets.     

Conceptual Modeling for Design Formulation

A design methodology is developed for the preliminary stage of building design. The methodology applies a theory of conceptual structure to model building design processes and incorporates the top-down approach of design decisions from a total system point of view. This design approach accounts for the uncertainties in the structural design processes and information, and provides a vehicle for formulating complex engineering decisions. The computing aspects of the design processes are categorised as arithmetic and reasoning formulations for capturing design information. The computing formulations are represented and manipulated, in the present study, using fuzzy logic, and their applications are illustrated with examples of building component design. The research results enable the implementation of a computer-aided tool for structural engineers to enhance their design decisions.     

Time-constrained reasoning under uncertainty

Dynamic classification problems present unique challenges beyond those of more traditionalstatic knowledge-based systems. Uncertain and incomplete input data, unpredictable event sequences, and critical time and resource constraints require new approaches and techniques for automated reasoning. Our work toward addressing these complex requirements has concentrated on developing an integrated software architecture which supports the knowledge engineering process from development to deployment. The approach we are using to deal with real-time issues in the deployment environment involves the use of a fast knowledge representation scheme, efficient forward and backward chaining mechanisms, and a meta-controller which handles asynchronous inputs, prioritized task requests, and hard performance deadlines.     

Extending information processing in a Fuzzy Random Forest ensemble

Imperfect information inevitably appears in real situations for a variety of reasons. Although efforts have been made to incorporate imperfect data into classification techniques, there are still many limitations as to the type of data, uncertainty, and imprecision that can be handled. In this paper, we will present a Fuzzy Random Forest ensemble for classification and show its ability to handle imperfect data into the learning and the classification phases. Then, we will describe the types of imperfect data it supports. We will devise an augmented ensemble that can operate with others type of imperfect data: crisp, missing, probabilistic uncertainty, and imprecise (fuzzy and crisp) values. Additionally, we will perform experiments with imperfect datasets created for this purpose and datasets used in other papers to show the advantage of being able to express the true nature of imperfect information.     

OFP_CLASS: a hybrid method to generate optimized fuzzy partitions for classification

The discretization of values plays a critical role in data mining and knowledge discovery. The representation of information through intervals is more concise and easier to understand at certain levels of knowledge than the representation by mean continuous values. In this paper, we propose a method for discretizing continuous attributes by means of fuzzy sets, which constitute a fuzzy partition of the domains of these attributes. This method carries out a fuzzy discretization of continuous attributes in two stages. A fuzzy decision tree is used in the first stage to propose an initial set of crisp intervals, while a genetic algorithm is used in the second stage to define the membership functions and the cardinality of the partitions. After defining the fuzzy partitions, we evaluate and compare them with previously existing ones in the literature.     

Evolutionary algorithms + domain knowledge = real-world evolutionary computation

We discuss implicit and explicit knowledge representation mechanisms for evolutionary algorithms (EAs). We also describe offline and online metaheuristics as examples of explicit methods to leverage this knowledge. We illustrate the benefits of this approach with four real-world applications. The first application is automated insurance underwriting-a discrete classification problem, which requires a careful tradeoff between the percentage of insurance applications handled by the classifier and its classification accuracy. The second application is flexible design and manufacturing-a combinatorial assignment problem, where we optimize design and manufacturing assignments with respect to time and cost of design and manufacturing for a given product. Both problems use metaheuristics as a way to encode domain knowledge. In the first application, the EA is used at the metalevel, while in the second application, the EA is the object-level problem solver. In both cases, the EAs use a single-valued fitness function that represents the required tradeoffs. The third application is a lamp spectrum optimization that is formulated as a multiobjective optimization problem. Using domain customized mutation operators, we obtain a well-sampled Pareto front showing all the nondominated solutions. The fourth application describes a scheduling problem for the maintenance tasks of a constellation of 25 low earth orbit satellites. The domain knowledge in this application is embedded in the design of a structured chromosome, a collection of time-value transformations to reflect static constraints, and a time-dependent penalty function to prevent schedule collisions.     

An Instance-Based Method for Remaining Useful Life Estimation for Aircraft Engines

Under customer service agreements (CSA), engine operational data are collected and stored for monitoring and analysis. Other data sources provide damage assessments that are either provided post-maintenance or analytically assessed. This paper takes advantage of these data and investigates local fuzzy models to determine the remaining useful life (RUL) of an engine or engine component. Local fuzzy models are related to both kernel regressions and locally weighted learning. The particular local models described in this paper are not based on individual models that consider the track history of a specific engine nor are they based on a global average model that would consider the collective track history of all the engines. Instead, for a given engine or component, this local fuzzy model defines a cluster of peers in which each of these peers is a similar instance to this given engine with comparable operational characteristics; the RUL prediction for this given engine is obtained by a fuzzy aggregation of its peers’ RUL. We combine the fuzzy instance-based approach with an evolutionary framework for model tuning and maintenance. This evolutionary tuning process is repeated periodically to automatically update and improve the fuzzy models such that they can be updated to date with the latest collection of data. This fuzzy instance-based approach is applied to predicting the RUL of a commercial engine validated with post-maintenance assessment. Reprinted with permission from Integration of Machinery Failure Prevention Technologies into Systems Health Management, Proceedings of the 61st Meeting of the Society for Machinery Failure Prevention Technology, Society for Machinery Failure Prevention Technology, 2007, on CD-ROM. This work was done while the author was with GE Global Research.