The automated information systems of support decision are the formation basis of the automated systems of monitoring, forecasting and management of natural processes in coastal sea regions.
All information on the processes happening in the region has to gather in the form of multi-purpose databases of indicators, parameters, criteria for evaluation of processes, calculation procedures and the analysis of the indicators grouped in spheres of activity; and knowledge bases of experts in various areas of ecological management; and database of standard decisions.
Structures of thematic databases (storages of data) on the main processes in environment of the coastal region, databases of indicators of these processes and criteria for evaluation of a situation on the main spheres of regional activity are proved in article.
Assuming that the credit is one of the most important banking products it follows that the quality assessment of customer creditworthiness is an essential factor for reducing the risk. With the intention to make a good assessment of creditworthiness many models and algorithms have been developed. Data mining algorithms for classification are very suitable for determining the validity of the application for credit. This paper presents an analysis of the effectiveness of the algorithms for classification of credit applications when they are used alone (as single classifier) as well as comparison with ensemble techniques usage. The techniques used as single classifiers are Neural Networks, Decision Trees and Support Vector Machines (SVM), and ensemble techniques AdaBoost and Bagging. K-fold cross-validation is used for model validation. Experiment is conducted in the Bosnian commercial bank dataset and results according to classification parameters such as accuracy and AUC are presented.
We present three methods to compute the expected value of the maximum of n independent, identically distributed exponentials and also obtain their Laplace Stieltjes Transform-LST. The results are applicable in the following cases: (1) the time to data loss in disk arrays with n-way replication, where the time to disk failure is exponentially distributed; (2) the time to completion of n exponentially distributed parallel tasks; (3) an upper bound to the mean fork-join response time when arrivals are Poisson, service times are exponentially distributed, so that response times are exponentially distributed.
The objective of the research is to expand the application area for the methodology of estimation of a production technical efficiency based on the stochastic frontier model. Some preliminary results show that the use of a precondition regarding the independence of random components of an error can lead to the estimates of technical efficiency that are (ranked) almost contrary to the true values under the condition that the true correlation coefficient for the components is close to 1.
In this article we consider a weaker degree of dependence between the random components of an error. In this situation correlation coefficient between the estimates obtained under assumption of independent random components (e.g. by using standard software) and the estimates obtained under assumption of dependent components increases from – 1 to 1. At the same time the estimates of technical efficiency provided by using of normal copula model are directly correlated with the true values. Thus the theory of copula functions can be adequately applied for the purpose of estimation of a production technical efficiency. The use of standard statistical software requires a verification of an assumption regarding independence of random components of an error. The absence of such an argumentation can lead to significant deviations between the estimated technical efficiencies and their true values.
It is also demonstrated that a confirmation of the hypothesis regarding the absence of inefficiency during the model’s parameters estimation then can be an evidence of a weak correlation between the true values of technical efficiencies and the estimates of technical efficiency obtained both by using standard software and by using theory of copula functions.
The work is devoted to developing methods for the automated processing of information based on specialized knowledge of the malicious actions (templates) and a knowledge base of regular situations and values their allowable characteristics, in order to detect intruders and malicious acts of automated data processing systems (ASPS).