TOPIC SEGMENTATION METHODS COMPARISON ON COMPUTER SCIENCE TEXTS

The demand for the creation of information systems that simplifies and accelerates work has greatly increased in the context of the rapidinformatization of society and all its branches. It provokes the emergence of more and more companies involved in the development of softwareproducts and information systems in general. In order to ensure the systematization, processing and use of this knowledge, knowledge managementsystems are used. One of the main tasks of IT companies is continuous training of personnel. This requires export of the content from the company'sknowledge management system to the learning management system. The main goal of the research is to choose an algorithm that allows solving theproblem of marking up the text of articles close to those used in knowledge management systems of IT companies. To achieve this goal, it is necessaryto compare various topic segmentation methods on a dataset with a computer science texts. Inspec is one such dataset used for keyword extraction andin this research it has been adapted to the structure of the datasets used for the topic segmentation problem. The TextTiling and TextSeg methods wereused for comparison on some well-known data science metrics and specific metrics that relate to the topic segmentation problem. A new generalizedmetric was also introduced to compare the results for the topic segmentation problem. All software implementations of the algorithms were written inPython programming language and represent a set of interrelated functions. Results were obtained showing the advantages of the Text Seg method incomparison with TextTiling when compared using classical data science metrics and special metrics developed for the topic segmentation task. Fromall the metrics, including the introduced one it can be concluded that the TextSeg algorithm performs better than the TextTiling algorithm on theadapted Inspec test data set.

USING OF MULTILAYER NEURAL NETWORKS FOR THE SOLVING SYSTEMS OF DIFFERENTIAL EQUATIONS

The article considers the study of methods for numerical solution of systems of differential equations using neural networks. To achieve this goal, thefollowing interdependent tasks were solved: an overview of industries that need to solve systems of differential equations, as well as implemented amethod of solving systems of differential equations using neural networks. It is shown that different types of systems of differential equations can besolved by a single method, which requires only the problem of loss function for optimization, which is directly created from differential equations anddoes not require solving equations for the highest derivative. The solution of differential equations’ system using a multilayer neural networks is thefunctions given in analytical form, which can be differentiated or integrated analytically. In the course of this work, an improved form of constructionof a test solution of systems of differential equations was found, which satisfies the initial conditions for construction, but has less impact on thesolution error at a distance from the initial conditions compared to the form of such solution. The way has also been found to modify the calculation ofthe loss function for cases when the solution process stops at the local minimum, which will be caused by the high dependence of the subsequentvalues of the functions on the accuracy of finding the previous values. Among the results, it can be noted that the solution of differential equations’system using artificial neural networks may be more accurate than classical numerical methods for solving differential equations, but usually takesmuch longer to achieve similar results on small problems. The main advantage of using neural networks to solve differential equations` system is thatthe solution is in analytical form and can be found not only for individual values of parameters of equations, but also for all values of parameters in alimited range of values.

THE USE OF MACHINE LEARNING METHODS FOR BINARY CLASSIFICATION OF THE WORKING CONDITION OF BEARINGS USING THE SIGNALS OF VIBRATION ACCELERATION

The paper investigates the relationship between vibration acceleration of bearings with their operational state. To determine these dependencies, a testbench was built and 112 experiments were carried out with different bearings: 100 bearings that developed an internal defect during operation and 12bearings without a defect. From the obtained records, a dataset was formed, which was used to build classifiers. Dataset is freely available. A methodfor classifying new and used bearings was proposed, which consists in searching for dependencies and regularities of the signal using descriptive functions: statistical, entropy, fractal dimensions and others. In addition to processing the signal itself, the frequency domain of the bearing operationsignal was also used to complement the feature space. The paper considered the possibility of generalizing the classification for its application on thosesignals that were not obtained in the course of laboratory experiments. An extraneous dataset was found in the public domain. This dataset was used todetermine how accurate a classifier was when it was trained and tested on significantly different signals. Training and validation were carried out usingthe bootstrapping method to eradicate the effect of randomness, given the small amount of training data available. To estimate the quality of theclassifiers, the F1-measure was used as the main metric due to the imbalance of the data sets. The following supervised machine learning methodswere chosen as classifier models: logistic regression, support vector machine, random forest, and K nearest neighbors. The results are presented in theform of plots of density distribution and diagrams.

DETERMINE RECOMMENDATION SYSTEMS TO SEARCH FOR BOOKS BY PREFERENCES OF WEB USERS

Currently, the question of state, formation and development of the information source interaction system, the scientific interaction and users' requestsin certain fields of activity remains relevant under the conditions of the development of the use of Internet services. Recommendation systems are oneof the types of artificial intelligence technologies for predicting parameters and capabilities.Due to the rapid increase in data on the Internet, it is becoming more difficult to find something really useful. And the recommendations offered by theservice itself may not always correspond to the user's preferences. The relevance of the topic is to develop a personal recommendation system forsearching books, which will not only reduce time and amount of unnecessary information, but also meet the user's preferences based on the analysis oftheir assessments and be able to provide the necessary information at the right time. All this makes resources based on referral mechanisms attractiveto the user. Such a system of recommendations will be of interest to producers and sellers of books, because it is an opportunity to provide personalrecommendations to customers according to their preferences.The paper considers algorithms for providing recommender systems (collaborative and content filtering systems) and their disadvantages.Combinations of these algorithms using a hybrid algorithm are also described. It is proposed to use a method that combines several hybrids in onesystem and consists of two elements: switching and feature strengthening. This made it possible to avoid problems arising from the use of each of thealgorithms separately.A literature web application was developed using Python using the Django and Bootstrap frameworks, as well as SQLite databases, and a system ofrecommendations was implemented to provide the most accurate suggestion. During the testing of the developed software, the work of the literatureservice was checked, which calculates personal recommendations for users using the method of hybrid filtering. The recommendation system wastested successfully and showed high efficiency.

SOFTWARE TESTING RESULTS ANALYSIS FOR THE REQUIREMENTS CONFORMITY USING NEURAL NETWORKS

The relevance of scientific work lies in the need to improve existing software designed to analyze the compliance of the results of software testing ofthe stated requirements. For the implementation of this goal, neural networks can be used by quality control specialists to make decisions aboutsoftware quality, or project managers as an expert system, for one of the quality indicators for the customer. The article deals with software testingwhich is a process of validation and verification of compliance of the software application or business program with the technical requirements thatguided its design and development, and work as expected, and identifies important errors or deficiencies classified by the severity of the program to befixed. Existing systems do not provide for or have only partial integration of systems of work with the analysis of requirements, which should ensurethe formation of expert assessment and provide an opportunity to justify the quality of the software product. Thus, a data processing model based on afuzzy neural network was proposed. An approach to allow determining the compliance of the developed software with functional and non-functionalrequirements was proposed, taking into account how successfully or unsuccessfully implemented this or that requirement. The ultimate goal ofscientific work is the development of algorithmic software analysis of compliance of software testing results to stated requirements for support in thedecisions taken. The following tasks are solved in scientific work: analysis of advantages and disadvantages of using existing systems when workingwith requirements; definition of general structure and classification of testing and requirements; characteristic main features of the use of neuralnetworks; designing architecture, the module of research of conformity of results of testing software to the stated requirements.

DYNAMICS OF AN ELECTRON BEAM FORMED BY MAGNETRON GUN WITH THE SECONDARY EMISSION CATHODE IN THE DECLINING MAGNETIC FIELD OF SOLENOID: EXPERIMENT AND SIMULATION

The article presents the results of research and calculations on the formation of a radial electron beam by a magnetron gun with a secondary emission cathode in the electron energy range 35...65 keV and measuring its parameters during transportation in the total decreasing magnetic field of thesolenoid and the stray field of permanent magnets. The beam was transported in a system consisting of copper rings with an inner diameter of 66 mm,located at a distance of 85 mm from the exit of the magnetron gun. The dependence of the beam current on the amplitude and gradient of the fielddecay has been studied. The studies carried out have shown the possibility of stable formation of a radial electron beam with an energy of tens of keVin the decreasing magnetic field of the solenoid. By optimizing the distribution of the magnetic field (created by the solenoid and ring magnets) and itsdecay gradient, it is possible to achieve an increase in the incident of electrons on one ring (up to ~72% of the beam current). On the basis of themathematical model of the movement of the electron flow, a software tool has been synthesized that makes it possible to obtain and interpret thecharacteristics of the resulting flows. The obtained numerical dependences are in satisfactory agreement with the experimental results for a magneticfield with a large decay gradient. Various configurations of the magnetic field are considered. Solutions to the direct problem of modeling electrontrajectories for given initial conditions and parameters are obtained. Various configurations of the magnetic field are considered. It is shown that forthe selected initial conditions for the electron beam and the distributions of the longitudinal magnetic field along the axis of the gun and the transportchannel, the electron flux falls on a vertical section, the length of which is on the order of a millimeter. Thus, by changing the amplitude anddistribution of the magnetic field, it is possible to control the current in the radial direction along the length of the pipe, and, therefore, the place of theelectron irradiation.

THE USE OF ARTIFICIAL INTELLIGENCE METHODS FOR APPROXIMATION OF THE MECHANICAL BEHAVIOR OF RUBBER-LIKE MATERIALS

In the XXI century, neural networks are widely used in various fields, including computer simulation and mechanics. This popularity is due to the factthat they give high precision, work fast and have a very wide range of settings. The purpose of creating a software product using elements of artificialintelligence, for interpolation and approximation of experimental data. The software should work correctly, and yield results with minimal error. Thedisadvantage of using mathematical approaches to calculating and predicting hysteresis loops is that they describe unloading rather poorly, thus, weobtain incorrect data for calculating the stress-strain state of a structure. The solution tool use of elements of artificial intelligence, but rather neuralnetworks of direct distribution. The neural network of direct distribution has been built and trained in this work. It has been trained with a teacher (ateacher using the method of reverse error propagation) based on a learning sample of a pre-experiment. Several networks of different structures werebuilt for testing, which received the same dataset that was not used during the training, but was known from the experiment, thus finding a networkerror in the amount of allocated energy and in the mean square deviation. The article describes in detail the mathematical interpretation of neuralnetworks, the method for training them, the previously conducted experiment, structure of network that was used and its topology, the training method,preparation of the training sample, and the test sample. As a result of the robots carried out, the software was tested in which an artificial neuralnetwork was used, several types of neural networks with different input data and internal structures were built and tested, the error of their work wasdetermined, the positive and negative sides of the networks that were used were formed.

COUNTERFACTUAL TEMPORAL MODEL OF CAUSAL RELATIONSHIPS FOR CONSTRUCTING EXPLANATIONS IN INTELLIGENT SYSTEMS

The subject of the research is the processes of constructing explanations based on causal relationships between states or actions of an intellectualsystem. An explanation is knowledge about the sequence of causes and effects that determine the process and result of an intelligent informationsystem. The aim of the work is to develop a counterfactual temporal model of cause-and-effect relationships as part of an explanation of the process offunctioning of an intelligent system in order to ensure the identification of causal dependencies based on the analysis of the logs of the behavior ofsuch a system. To achieve the stated goals, the following tasks are solved: determination of the temporal properties of the counterfactual description ofcause-and-effect relationships between actions or states of an intelligent information system; development of a temporal model of causal connections,taking into account both the facts of occurrence of events in the intellectual system, and the possibility of occurrence of events that do not affect theformation of the current decision. Conclusions. The structuring of the temporal properties of causal links for pairs of events that occur sequentially intime or have intermediate events is performed. Such relationships are represented by alternative causal relationships using the temporal operators"Next" and "Future", which allows realizing a counterfactual approach to the representation of causality. A counterfactual temporal model of causalrelationships is proposed, which determines deterministic causal relationships for pairs of consecutive events and pairs of events between which thereare other events, which determines the transitivity property of such dependencies and, accordingly, creates conditions for describing the sequence ofcauses and effects as part of the explanation in intelligent system with a given degree of detail The model provides the ability to determine cause-andeffect relationships, between which there are intermediate events that do not affect the final result of the intelligent information system.

INFORMATION TECHNOLOGY OF FORMING OPTION FOR LOGISTICS DISTRIBUTION CHANNEL CONFIGURATION RESISTANT TO EMERGENCIES

The problem and the main stages of choosing a rational configuration of a four-level logistics network that is resistant to the impact of emergencies forstrategic planning are considered. The problem under consideration belongs to the class of multicriteria optimization problems. Criteria related to thefinancial costs of building and operating a logistics distribution channel, as well as criteria related to the level of quality of customer service, arecontradictory. To solve the problem of stability of the logistics system configuration to emergencies, such as failure of intermediate warehouses,failure of transport arteries, etc., a strategic management information system was developed by integrating existing software components at the level ofenterprise software applications. The integration of the system was based on a service-oriented architecture, as all its components are heterogeneous innature. This approach allows you to reuse existing program code. To determine a sustainable configuration option, two criteria are used, which areconsidered equivalent: the level of costs for the maintenance of the logistics channel and the level of service quality in the event of differentemergencies. Since the probability of emergencies is unknown, the minimax criterion is used to minimize the risk when choosing a rationalconfiguration of the logistics network. For this purpose, losses from emergencies are calculated according to all criteria, and there is a variant of thelogistics network configuration that will be the least risky. That is, we will not be able to get a worse result than the one we rely on. The results of thestudy are presented in the form of a configuration variant of the logistics distribution system, which can be used in the future to determine businessoptions.

ESTIMATING WITH A GIVEN ACCURACY OF THE COEFFICIENTS AT NONLINEAR TERMS OF UNIVARIATE POLYNOMIAL REGRESSION USING A SMALL NUMBER OF TESTS IN AN ARBITRARY LIMITED ACTIVE EXPERIMENT

We substantiate the structure of the efficient numerical axis segment an active experiment on which allows finding estimates of the coefficients fornonlinear terms of univariate polynomial regression with high accuracy using normalized orthogonal Forsyth polynomials with a sufficiently smallnumber of experiments. For the case when an active experiment can be executed on a numerical axis segment that does not satisfy these conditions, wesubstantiate the possibility of conducting a virtual active experiment on an efficient interval of the numerical axis. According to the results of the experiment, we find estimates for nonlinear terms of the univariate polynomial regression under research as a solution of a linear equalities system withan upper non-degenerate triangular matrix of constraints. Thus, to solve the problem of estimating the coefficients for nonlinear terms of univariatepolynomial regression, it is necessary to choose an efficient interval of the numerical axis, set the minimum required number of values of the scalarvariable which belong to this segment and guarantee a given value of the variance of estimates for nonlinear terms of univariate polynomial regressionusing normalized orthogonal polynomials of Forsythe. Next, it is necessary to find with sufficient accuracy all the coefficients of the normalized orthogonal polynomials of Forsythe for the given values of the scalar variable. The resulting set of normalized orthogonal polynomials of Forsythe allows us to estimate with a given accuracy the coefficients of nonlinear terms of univariate polynomial regression in an arbitrary limited active experiment: the range of the scalar variable values can be an arbitrary segment of the numerical axis. We propose to find an estimate of the constant and ofthe coefficient at the linear term of univariate polynomial regression by solving the linear univariate regression problem using ordinary least squaresmethod in active experiment conditions. Author and his students shown in previous publications that the estimation of the coefficients for nonlinearterms of multivariate polynomial regression is reduced to the sequential construction of univariate regressions and the solution of the correspondingsystems of linear equalities. Thus, the results of the paper qualitatively increase the efficiency of finding estimates of the coefficients for nonlinearterms of multivariate polynomial regression given by a redundant representation.