Data mining techniques frequently find a large number of patterns or rules, which make it very difficult for a human analyst to interpret the results and to find the truly interesting and actionable rules. Due to the subjective nature of "interestingness", human involvement in the analysis process is crucial. In this paper, we propose a novel visual data mining framework for the purpose of identifying actionable knowledge quickly and easily from discovered rules and data. This framework is called the Opportunity Map. It is inspired by some interesting ideas from Quality Engineering, in particular Quality Function Deployment (QFD) and the House of Quality. It associates summarized data or discovered rules with the application objective using an interactive matrix, which enables the user to quickly identify where the opportunities are. The proposed system can be used to visually analyze discovered rules, and other statistical properties of the data. The user can also interactively group actionable attributes and values, and see how they affect the targets of interest. Combined with drill-down and comparative analysis, the user can analyze rules and data at different levels of detail. The proposed visualization framework thus represents a systematic and yet flexible method of rule analysis. Applications of the system to large-scale data sets from our industrial partner have yielded promising results.

Data mining algorithms usually generate a large number of rules, which may not always be useful to human users. In this project, we propose a novel visual data-mining framework, called Opportunity Map, to identify useful and actionable knowledge quickly and easily from the discovered rules. The framework is inspired by the House of Quality from Quality Function Deployment (QFD) in Quality Engineering. It associates discovered rules, related summarized data and data distributions with the application objective using an interactive matrix. Combined with drill down visualization, integrated visualization of data distribution bars and rules, visualization of trend behaviors, and comparative analysis, the Opportunity Map allows users to analyze rules and data at different levels of detail and quickly identify the actionable knowledge and opportunities. The proposed framework represents a systematic and flexible approach to rule analysis. Applications of the system to large-scale data sets from our industrial partner have yielded promising results.

Pinpoint is a promising first step towards using a rich model of task context in proactive and dynamic IR systems. Pinpoint allows a user to navigate decision tree representations of problem spaces, built by domain experts, while dynamically entering annotations specific to their problem. The system then automatically generates queries to information repositories based on both the user's annotations and location in the problem space, producing results that are both task focused and problem specific. Initial feedback from users and domain experts has been positive.

Visualization of a search process can be an effective way to verify the performance of a search algorithm, especially in terms of its coverage of the search space and behavior near local optima. Planar and 3D surface graphs provide the best intuition, e.g., for highlighting "nearest neighbor" solutions. However, they cannot be directly applied to most search and optimization problems because of the high dimensionality of the search space. Building on insights from space-filling curves and their application for combinatorial optimization problems, This work presents a technique for mapping the solution space of hundred-variable combinatorial optimization problems into two-dimensions. Experimental results for electronics assembly optimization are presented as an example of this technique. Mathematical properties of the proposed N-to-2-space transformation are discussed, and several sample visualizations are presented.

Analyzing data to find trends, correlations, and stable patterns is an important task in many industrial applications. This paper proposes a new technique based on parallel coordinate visualization. Previous work on parallel coordinate methods has shown that they are effective only when variables that are correlated and/or show similar patterns are displayed adjacently. Although current parallel coordinate tools allow the user to manually rearrange the order of variables, this process is very time-consuming when the number of variables is large. Automated assistance is required. This paper introduces an edit-distance based technique to rearrange variables so that interesting change patterns can be easily detected visually. The Visual Miner (V-Miner) software includes both automated methods for visualizing common patterns and a query tool that enables the user to describe specific target patterns to be mined or displayed by the system. In addition, the system can filter data according to rules sets imported from other data mining tools. This feature was found very helpful in practice, because it enables decision makers to visually identify interesting rules and data segments for further analysis or data mining. This paper begins with an introduction to the proposed techniques and the V-Miner system. Next, a case study illustrates how V-Miner has been used at Motorola to guide product design and test decisions.

Analyzing data to find trends, correlations, and stable patterns is an important problem for many industrial applications. We propose a new technique based on parallel coordinates visualization. Previous work on parallel coordinates method has shown that they are effective only when variables that are correlated and/or show similar patterns are displayed adjacently. Although current parallel coordinates tools allow the user to manually rearrange the order of variables, this process is very time-consuming when the number of variables is large. Automated assistance is needed. We propose an edit-distance based technique to rearrange variables so that interesting patterns can be easily detected. Our system, V-Miner, includes both automated methods for visualizing common patterns and a query tool that enables the user to describe specific target patterns to be mined/displayed by the system. Following an overview of the system, a case study is presented to explain how Motorola engineers have used V-Miner to identify significant patterns in their product test and design data.

Classification is one of the fundamental tasks of data mining. Most rule induction and decision tree algorithms perform a local, greedy search to generate classification rules that are often more complex than necessary. Evolutionary algorithms for pattern classification have recently received increased attention because they can perform global searches. In this paper, we propose a new approach for discovering classification rules by using gene expression programming (GEP), a new technique of genetic programming (GP) with linear representation. The antecedent of discovered rules may involve many different combinations of attributes. To guide the search process, we suggest a fitness function considering both the rule consistency gain and completeness. A multiclass classification problem is formulated as multiple two-class problems by using the one-against-all learning method. The covering strategy is applied to learn multiple rules if applicable for each class. Compact rule sets are subsequently evolved using a two-phase pruning method based on the minimum description length (MDL) principle and the integration theory. Our approach is also noise tolerant and able to deal with both numeric and nominal attributes. Experiments with several benchmark data sets have shown up to 20% improvement in validation accuracy, compared with C4.5 algorithms. Furthermore, the proposed GEP approach is more efficient and tends to generate shorter solutions compared with canonical tree-based GP classifiers.

Applications of computer simulation for surface mount technology (SMT) assembly lines have addressed many aspects of supply chain management, e.g., estimating cycle times, evaluating production equipment and line configurations, optimizing throughput and assessing product design for manufacturability. The implementation of an SMT machine simulator typically requires specialized code based on cycle time characterization experiments for the given machine type. This paper describes a generic simulation toolkit based on a classification of current SMT equipment. A detailed survey of machines has led to the definition of machine families and a generic machine model. A simulation toolkit consisting of object classes written in C++ has been developed to represent the generic machine model. Object-oriented design has enabled the construction of an object hierarchy that models the physical structure of a machine and its operation. A machine definition language has been specified to let the user configure the toolkit for a particular machine. The concept of a "snapshot" object is introduced, as a way to capture simulated operations for both parallel and sequential processes. Finally, this paper presents an application of the toolkit in the optimization of SMT machines. Results are presented for the Universal GSM2, a gantry-type machine.

This paper introduces an optimizer for a new family of modular, multistation, walking beam, high-speed chip mounters. The objective is to optimize the machines in a manner that would streamline the use of nozzles and part feeder mechanisms and at the same time increase throughput. The optimization of these machines is a large NP-complete problem, and therefore, a heuristic search method is needed to solve the problem in reasonable time. Four knowledge-based systems are introduced to solve this problem. These systems were designed to emulate human experts, who have optimized these types of machines manually. Benchmarks were performed for 18 industrial test cases. The results show that overall, the knowledge-based systems outperformed software supplied by the vendor of the machine in both feeder slot savings and throughput. This performance represents a key improvement, and a prototype system has been implemented in our industrial partner's factory.

Cast in a fractal architecture, the model of FMS admits a natural hierarchical decomposition of highly decoupled units with similar structure and control. As a consequence, this model manages the structural complexity and coordination of an FMS hierarchy by maximizing local functionality and minimizing global control. Most significantly, the software that governs the behavior of the basic fractal unit, from which this architecture is built, is based on a common template that allows for a high degree of abstraction, object orientation, and reuse across the hierarchy.>