Leonid Sheremetov

Development of intelligent reusable learning objects for web-based education systems

The research in WBE (Web-based Education) systems is centered in reusability, accessibility, durability and interoperability of didactic materials and environments of virtual education. A special type of labeled materials called Intelligent Reusable Learning Components Object Oriented (IRLCOO), producing learning materials with interface and functionality standardized, rich in multimedia, interactivity and feedback is described in this paper. The structuring model for dynamic composition of these components is based on the concept graph knowledge representation model. The multi-agent architecture as a middleware for open WBE systems is developed for sequencing and delivery of learning materials composed of IRLCOOs.

Soft-computing technologies for configuration of cooperative supply chain

Abstract In this article, the problem of dynamic configuration of cooperative supply chain (CSC) as a dynamic, flexible and agile system is considered. Members of the CSC negotiate and compromise on the optimal configuration in a spirit of Cupertino, in order to meet commitments made to each other. Based on this framework, the proposed approach considers configuring as: (i) coalition formation and (ii) product and resource allocation tasks in multi-agent environment. To solve the configuring task, an application of different techniques of soft computing is examined. The first approach permits to find sub-optimal solution applying the theory of games with fuzzy coalitions. In this case, genetic algorithms (GA) are used to find solutions of the game. The second approach uses: (i) genetic algorithms directly and (ii) constraint satisfaction problem solving for resource allocation task. Ontologies codified as object oriented constraint networks are used for task description and decomposition. A multi-agent test-bed based on FIPA compliant agent platform is developed and used to conduct the experiments. The above techniques are compared and the simulation results are discussed.

Perception-Based Functions in Qualitative Forecasting

Summary. Perception-based function (PBF) is a fuzzy function obtained as a result of reconstruction of human judgments given by a sequence of rules Rk: If T is Tk then S is Sk, where Tk are perception-based intervals defined on the domain of independent variable T, and Sk are perception-based shape patterns of variable S on interval Tk. Intervals Tk can be expressed by words like Between N and M, Approximately M, Middle of the Day, End of the Week, etc. Shape patterns Sk can be expressed linguistically, e.g., as follows: Very Large, Increasing, Quickly Decreasing and Slightly Concave, etc. PBF differs from the Mamdani fuzzy model which defines a crisp function usually obtained as a result of tuning of function parameters in the presence of training crisp data. PBF is used for reconstruction of human judgments when testing data are absent or scarce. Such reconstruction is based mainly on scaling and granulation of human knowledge. PBF can be used in Computing with Words and Perceptions for qualitative evaluation of relations between variables. In this chapter we discuss application of PBF to qualitative forecasting of a new product life cycle. We consider new parametric patterns used for modeling convex–concave shapes of PBF and propose a method of reconstruction of PBF with these shape patterns. These patterns can be used also for time series segmentation in perception-based time series data mining.

Intelligent multi-agent support for the contingency management system

Abstract In the present paper, we describe agent-based intelligent infrastructure of contingency management system (CMS). This infrastructure supports information collection from distributed heterogeneous databases, integration with enterprise legacy software systems, logistics planning, and monitoring of contingency situation in open, dynamic agent environment. The approach is applied for the development of the CMS for the oil complexes in the marine zone of the gulf of Campeche. One of the key components of the CMS, an evacuation logistics planning system is described to illustrate the approach. A method of coalition formation is used for logistics planning in a multi-agent environment. Distributed and centralized algorithms for coalition formation are described and compared. The latter algorithm is based on the cooperative game theoretic model with fuzzy coalitions. In both algorithms, agents use fuzzy rules of Mamdani type for decision making. Implementation issues are discussed.

Design and implementation of a FIPA compliant agent platform in .NET

The aim of this paper is to describe the design and implementation of an agent platform called CAPNET (Component Agent Platform based on .NET) that is fully compliant with the specifications of the Foundation for Intelligent Physical Agents (FIPA) and implemented as 100% managed code in the .NET framework. The tools for the platform configuration and administration are presented and a case study is provided that shows the usability of the CAPNET in industrial and mission critical scenarios.

Multilayer Neural Network with Multi-Valued Neurons in time series forecasting of oil production

In this paper, we discuss the long-term time series forecasting using a Multilayer Neural Network with Multi-Valued Neurons (MLMVN). This is a complex-valued neural network with a derivative-free backpropagation learning algorithm. We evaluate the proposed approach using a real-world data set describing the dynamic behavior of an oilfield asset located in the coastal swamps of the Gulf of Mexico. We show that MLMVN can be efficiently applied to univariate and multivariate one-step- and multi-step ahead prediction of reservoir dynamics. This paper is not only intended for proposing to use a complex-valued neural network for forecasting, but to deeper study some important aspects of the application of ANN models to time series forecasting that could be of the particular interest for pattern recognition community. A long-term forecasting of oil production was done using MLMVN.Univariate and multivariate forecasting models were developed.MLMVN is efficient for both multivariate and univariate forecasting models.MLMVN-based prediction combines both regression and pattern recognition approaches.

Fuzzy expert system for solving lost circulation problem

Lost circulation is the most common problem encountered when drilling. This paper describes a distributed hybrid intelligent system, called SmartDrill, using fuzzy logic, expert system framework and Web services for helping petroleum engineers to diagnose and solve lost circulation problems. The fuzzy algebra of strict monotonic operations is used as an underlying model for expert system development. Its realization in inference procedures of expert systems is simpler than for expert systems based on lexicographic operations. Overall, the system architecture is discussed and implementation details are provided. The system is aimed to help in making decisions at the operational level and is at field testing phase in PEMEX, Mexican Oil Company.

Moving Approximation Transform and Local Trend Associations inTime Series Data Bases

Summary. The properties of moving approximation (MAP) transform and its application to time series data mining are discussed. MAP transform replaces time series values by slope values of lines approximating time series data in sliding window. A simple method of MAP transform calculation for time series with fixed time step is proposed. Based on MAP the measures of local trend associations and local trend distances are introduced. These measures are invariant under independent linear transformations and normalizations of time series values. Measure of local trend associations defines association function and measure of association between time series. The methods of application of association measure to construction of association network of time series and clustering are proposed and illustrated by examples of economic, financial, and synthetic time series.

A novel associative model for time series data mining

We introduce a novel non-linear forecasting technique based on the Gamma classifier.Its performance for long-term time horizons was tested on synthetic and real data.Two benchmark time series were used for testing.Six time series related to monthly oil production were also used.The Gamma classifier model outperformed previous techniques in forecast accuracy. The paper describes a novel associative model for time series data mining. The model is based on the Gamma classifier, which is inspired on the Alpha-Beta associative memories, which are both supervised pattern recognition models. The objective is to mine known patterns in the time series in order to forecast unknown values, with the distinctive characteristic that said unknown values may be towards the future or the past of known samples. The proposed model performance is tested both on time series forecasting benchmarks and a data set of oil monthly production. Some features of interest in the experimental data sets are spikes, abrupt changes and frequent discontinuities, which considerably decrease the precision of traditional forecasting methods. As experimental results show, this classifier-based predictor exhibits competitive performance. The advantages and limitations of the model, as well as lines of improvement, are discussed.

Towards Perception Based Time Series Data Mining

Human decision making procedures in problems related with analysis of time series data bases (TSDB) often use perceptions like “several days”, “high price”, “quickly increasing” etc. Computing with Words and Perceptions can be used to formalize perception based expert knowledge and inference mechanisms defined on numerical domains of TSDB. For extraction from TSDB perception based information relevant to decision making problems it is necessary to develop methods of perception based time series data mining (PTSDM). The paper considers different approaches used in analysis of time series databases for description of perception based patterns and discusses some methods of PTSDM.