Kevin McCarty

Contextual fuzzy type-2 hierarchies for decision trees (CoFuH-DT) — An accelerated data mining technique

Advanced data mining techniques (ADMT) are very powerful tools for classification, understanding and prediction of object behaviors, providing descriptive relationships between objects such as a customer and a product they intend to buy. ADMT typically consists of classifiers and association techniques, among them, decision trees (DT). However, some important relationships are not readily apparent in a traditional decision tree. In addition, decision trees can grow quite large as the number of dimensions and their corresponding elements increase, requiring significant resources for processing. In either case, rules governing these relationships can be difficult to construct. This paper presents CoFuH-DT, a new algorithm for capturing intrinsic relationships among the nodes of DT, based upon a proposed concept of type-2 fuzzy ldquocontextsrdquo. This algorithm modifies a decision tree, first by generating type-1 fuzzy extensions of the underlying DT criteria or ldquoconditionsrdquo; combining further those extensions into new abstractions overlaid with type-2 contexts. The resulting fuzzy type-2 classification is then able to capture intrinsic relationships that are otherwise non-intuitive. In addition, performing fuzzy setbased operations simplifies the decision tree much faster than traditional search techniques in order to aid in rule construction. Testing presented on a virtual store example demonstrates savings of multiple orders of magnitude in terms of nodes and applicable conditions resulting in 1) reduced complexity of decision tree, 2) ability to data mine difficult to detect interrelationships, 3) substantial acceleration of decision tree search, making it applicable for 4) real-time data mining of new knowledge.

A temporal-spatial data fusion architecture for monitoring complex systems

Non-homogenous systems arise from the need to incorporate a variety of disparate systems into a cohesive functioning whole and may comprise many crucial elements of an industrialized, modern society. As a result they must be constantly monitored to ensure efficient functioning and avoid expensive breakdowns. In particular, inter-connected computer-based systems must increasingly be aware of cyber and physical threats that are dynamic and evolutionary in nature. However, difficulties arise in trying to ascertain threats and problems among the diverse sources of information generated by these systems. Finally, there is the question of how best to present this data to a human operator. Human systems require not just analysis, but presentation which encourages timely, proactive or corrective decisions. This paper presents a software architecture to solve these problems based upon data fusion using temporal-spatial relationships. As phase one of a three phase project, a prototype implementation of this architecture demonstrates application of this technique for a cohesive system. Test results showed the system capable of real-time fusion of physical, cyber and process data elements as well as analysis, display and interpretation of threats.

Adaptive behavioral control of collaborative robots in hazardous environments

Terrain exploration carries with it significant hazards. Robots attempting to map a piece of unknown terrain must be able to make decisions and react appropriately to dynamic and potentially hostile conditions. However, because of constraints on size and cost, robots may have limited ability to store and process necessary information. In addition, knowledge discovered by others may be difficult to share. This paper proposes a system using a powerful master controller, operating from a safe environment, directing the movements of numerous robots exploring a piece of terrain. The master controller processes the information from the robots, updates the decision process and distributes these updates back to the robots. This process allows for a cooperative, effective search environment while also maintaining a small processing footprint. It also allows the robot to employ adaptive, subsumptive behavioral modification as new information is made available. A test simulation of a hazardous environment demonstrates that even robots with little intrinsic intelligence can learn complex behaviors in order to reach their goal.

Line-of-sight tracking based upon modern heuristics approach

Any autonomous vehicle must be able to successfully navigate a wide variety of situations and terrain conditions. As a result, proposed solutions usually involve a sophisticated and expensive implementation of both hardware and software. In many situations, however, truly autonomous operation may not be necessary or practical. Instead, equipping and training a vehicle to automatically follow a human-controlled lead vehicle is a viable alternative. While still autonomous, the vehicle relies upon its leader to handle the complex decisions with regards to course and speed. This paper presents a simple and elegant configuration, called FLoST for fuzzy line of sight tracking, based on inexpensive line-of-sight devices controlled by a heuristic to determine direction and speed of a follower. Unlike the alternative approach where the follower needs to undergo a complex training process, the follower using the approach presented in this paper primarily relies upon a human leader to provide direction, allowing for a much simpler and less expensive vehicle implementation while still being able to match or exceed the effectiveness of the autonomous design under similar conditions. Finally, three boundary cases of lead vehicle maneuvers (circle, spiral and weave) are presented to show the efficacy of this approach.

A fuzzy framework with modeling language for type 1 and type 2 application development

Fuzzy logic, Type-1 and Type-2, is well suited for human systems interactions because they provides a natural way of implementing linguistic terms from human experts. Existing fuzzy frameworks, however, provide limited support for Type-2. They also tend to be fairly complicated and/or have limited portability. This paper introduces a fuzzy framework for building a Type-1Type-2 fuzzy controller. A “wizard” application and modeling language are supported to provide an easy-to-use interface for creating a fuzzy inference system. The benefits of this framework are: (1) Increased understanding of fuzzy systems implementation via easy-to-use visual tools; (2) Reduced development time; (3) A standardized and portable codebase; (4) Easy configuration via XML; (5) Support for both Type-1 and Type-2 fuzzy sets and rules. The framework is tested and solves a maze problem using both Type-1 and Type-2 implementations.

A database driven memetic algorithm for fuzzy set optimization

Fuzzy logic provides a natural and precise way for humans to define and interact with systems. Optimizing a fuzzy inference system, however, presents some special challenges for the developer because of the imprecision that is inherent to fuzzy sets. This paper expands upon an earlier development of a fuzzy framework, adding components for dynamic self-optimization. What makes this approach unique is the use of relational database as a computational engine for the memetic algorithm and fitness function. The new architecture combines the power of fuzzy logic with the special properties of a relational database to create an efficient, flexible and self-optimizing combination. Database objects provide the fitness function, population sampling, gene crossover and mutation components allowing for superior batch processing and data mining potential. Results show the framework is able to improve the performance of a working configuration as well as fix a non-working configuration.

Fuzzy Contexts (Type C) and fuzzymorphism to solve situational discontinuity problems

Generalized solutions to complex problems often suffer from being overly complicated. The main contribution of this paper is to describe an architecture that allows for greater problem generalization without the traditional corresponding increase in complexity. The architecture extends traditional fuzzy logic and is called Fuzzy Contexts or Fuzzy Logic Type-C. Fuzzy logic permits partial membership and values can belong to multiple fuzzy sets. By breaking down a problem space into smaller contexts and allowing algorithms themselves to have relaxed memberships in those contexts, a Type-C solution can support multiple solutions to complex problems. This paper describes how problem spaces may be decomposed into smaller, more easily solvable components and fuzzified together under a Type-C hierarchy. Test results with a simulated robotic navigation system demonstrates how a Type-C implementation is able to improve upon a generalized fuzzy controller.

A proposed data fusion architecture for micro-zone analysis and data mining

Micro-zone analysis involves use of data fusion and data mining techniques in order to understand the relative impact of many different variables. Data Fusion requires the ability to combine or “fuse” date from multiple data sources. Data mining involves the application of sophisticated algorithms such as Neural Networks and Decision Trees, to describe micro-zone behavior and predict future values based upon past values. One of the difficulties encountered in developing generic time series or other data mining techniques for micro-zone analysis is the wide variability of the data sets available for analysis. This presents challenges all the way from the data gathering stage to results presentation. This paper presents an architecture designed and used to facilitate the collection of disparate data sets well suited for data fusion and data mining. Results show this architecture provides a flexible, dynamic framework for the capture and storage of a myriad of dissimilar data sets and can serve as a foundation from which to build a complete data fusion architecture.

Contextual Data Rule Generation For Autonomous Vehicle Control

Autonomous vehicles are often called upon to deal with complex and varied situations. This requires analyzing input from sensor arrays to get as accurate a description of the environment as possible. These ad-hoc descriptions are then compared against existing rule sets generated from decision trees that decide upon a course of action. However, with so many environmental conditions it is often difficult to create decision trees that can account for every possible situation, so techniques to limit the size of the decision tree are used. Unfortunately, this can obscure data which is sparse, but also important to the decision process. This paper presents an algorithm to analyze a decision tree and develops a set of metrics to determine whether or not sparse data is relevant and should be include. An example demonstrating the use of this technique is shown.

Descending Deviation Optimization techniques for scheduling problems

In factory automation, production line scheduling entails a number of competing issues. Finding optimal configurations often requires use of local search techniques. Local search looks for a goal state employing heuristics and random local ldquoprobesrdquo in order to move from state to state. All local search techniques, however, suffer from problems with local maxima, i.e. have the potential of getting ldquostuckrdquo in a suboptimal state. While careful introduction of randomizations is certainly a recognized technique, it can also lead the algorithm even more astray. This paper describes a heuristic technique called descending deviation optimizations (DDO) in which a gradually lowering-randomization ceiling allows a local search technique to ldquobouncerdquo randomly without going too far astray. An example applying the DDO to a local search technique and achieving significant improvement is shown.