Normative expert systems have not become commonplace because they have been difficult to build and use. Over the past decade, however, researchers have developed the influence diagram, a graphical representation of a decision maker's beliefs, alternatives, and preferences that serves as the knowledge base of a normative expert system. Most people who have seen the representation find it intuitive and easy to use. Consequently, the influence diagram has overcome significantly the barriers to constructing normative expert systems. Nevertheless, building influence diagrams is not practical for extremely large and complex domains. In this book, I address the difficulties associated with the construction of the probabilistic portion of an influence diagram, called a knowledge map, belief network, or Bayesian network. I introduce two representations that facilitate the generation of large knowledge maps. In particular, I introduce the similarity network, a tool for building the network structure of a knowledge map, and the partition, a tool for assessing the probabilities associated with a knowledge map. I then use these representations to build Pathfinder, a large normative expert system for the diagnosis of lymph-node diseases (the domain contains over 60 diseases and over 100 disease findings). In an early version of the system, I encoded the knowledge of the expert using an erroneous assumption that all disease findings were independent, given each disease. When the expert and I attempted to build a more accurate knowledge map for the domain that would capture the dependencies among the disease findings, we failed. Using a similarity network, however, we built the knowledge-map structure for the entire domain in approximately 40 hours. Furthermore, the partition representation reduced the number of probability assessments required by the expert from 75,000 to 14,000.

Probabilistic Similarity Networks

The success rate of software projects has been increased since agile methodologies were adopted by many companies. Due their flexibility and continuous communication with clients, the main reason for the failure has shifted from the formulation and understanding of the requirements to inaccurate effort estimation. In recent years, several researchers and practitioners have proposed different estimation techniques. However, some projects are still failing because the budget and/or schedule are not accurately estimated since there still are numerous uncertain variables in software development process. Previous team collaborations, expertise and experience of team members, frequency of changing requirements or priorities are just a few examples. To improve the accuracy of effort estimation, this research proposes a model for agile software development project prediction using Bayesian networks. Based on literature review and practitioners’ knowledge, we identified two major categories of factors that influence effort needed: teamwork quality and user stories characteristics. We identified the sub-factors for each category and inter-dependencies between them. In our model, these factors are the nodes of the directed acyclic graph. The model can help agile teams to obtain a better software effort estimation.

Effort Prediction in Agile Software Development with Bayesian Networks

Planning and forecasting in sport is an integral element of the sphere and system of physical culture and sport functioning, and many related areas of activity. The development of information technologies, methods and means of artificial intelligence gives wide opportunities for their application in the field of sports forecasting. The article describes the features of design and application of the system that predicts sports results in team and individual sports. The functional structure of the system is given, the basic principles of its operation are considered. A feature of the structure is the introduction of a data clustering module based on a neural network of vector quantization of signals. The algorithm of this module functioning and the possibility of its expansion for detailing the formed forecast is described. The recommendations to users on the formation and detailing of the overall sample, evaluation of the generated sample effectiveness and analysis of the results are defined. The results of experiments on synthetic samples and real methods are presented and analyzed. Convenient and intuitive interface allows the system to be used by specialists of different profiles in the field of physical culture and sports, starting with the athletes themselves and ending with the heads of sports clubs, organizations and federations. Keywords—sports forecasting; software system; neural network; training sample; vector quantization; LVQ-network

Neural Network of Vector Quantization as a Tool for Predicting Sports Achievements in Individual and Team Sports

Bayesian networks (BN) are a key information technology for dealing with probabilities in artificial intelligence. In the present work we introduce an application of BN as a tool for estimating injury risk in high level rhythmic gymnastics. At this stage we propose the base structure of the model consisting of five subnets, contributing to overall injury risk. Most of the model conditional probability tables are estimated with T Normal functions – a featu r e included in Agena Risk tool. Sensitivity analysis characterizes the degree of influence of the different input factors, which is consistent with expert knowledge. The model results are satisfactory for the test set of gymnasts in the current competitive season. Quantitative predictions show a s ignificant opportunity for reducing injury rate, but further data collection and research are necessary to improve the precision of the model.

Kristina Petkova

Papers

Bayesian Network - Based Causal Analysis of Injury Risk in Elite Rhythmic Gymnastics

Analysis and assessment of injury risk in female gymnastics:Bayesian Network approach