Radio Electronics, Computer Science, Control

COMPARISON OF SHORT-TERM FORECASTING METHODS OF ELECTRICITY CONSUMPTION IN MICROGRIDS

2023-02-24T14:20:47+02:00

Context. The current stage of development of the electric power industry is characterized by an intensive process of microgrid development and management. The feasibility of using a microgrid is determined by the fact that it has a number of advantages compared to classical methods of energy generation, transmission, and distribution. It is much easier to ensure the reliability of electricity supply within the microgrid than in large energy systems. Energy consumers in a microgrid can affect the power balancing process by regulating their loads, generating, storing, and releasing electricity. One of the main tasks of the microgrid is to provide consumers with electrical energy in a balance between its generation and consumption. This is achieved thanks to the intelligent management of the microgrid operation, which uses energy consumption forecasting data. This allows to increase the efficiency of energy infrastructure management.

Objective. The purpose of this work is to develop short-term electricity consumption forecasting models for various types of microgrid electricity consumers, which will improve the efficiency of energy infrastructure management and reduce electricity consumption.

Method. The SARIMA autoregressive model and the LSTM machine learning model are used to obtain forecast values of electricity consumption. AIC and BIC information criteria are used to compare autoregressive models. The accuracy of forecasting models is evaluated using MAE, RMSE, MAPE errors.

Results. The experiments that forecast the amount of electricity consumption for the different types of consumers were conducted. Forecasting was carried out for both LSTM and AR models on formed data sets at intervals of 6 hours, 1 day, and 3 days. The forecasting results of the LSTM model met the forecasting requirements, providing better forecasting quality compared to AR models.

Conclusions. The conducted study of electricity consumption forecasting made it possible to find universal forecasting models that meet the requirements of forecasting quality. A comparative analysis of developed time series forecasting models was performed, as a result of which the advantages of ML models over AR models were revealed. The predictive quality of the LSTM model showed the accuracy of the MAPE of forecasting electricity consumption for a private house – 0.1%, a dairy plant – 3.74%, and a gas station – 3.67%. The obtained results will allow to increase the efficiency of microgrid management, the distribution of electricity between electricity consumers to reduce the amount of energy consumption and prevent peak loads on the power grid.

METHOD OF ROUTING A GROUP OF MOBILE ROBOTS IN A FIXED NETWORK FOR SEARCHING THE MISSING OBJECTS IN A TECHNOLOGICAL DISASTER ZONE

2023-02-27T18:59:18+02:00

Context. The relevance of the article is determined by the need for further development of models of collective behavior of systems with multi-agent structure construction endowed with intelligence that ensures synchronization of the joint efforts of various agents while achieving the goals set for the system. The method proposed in the article solves the problem of competition between different agents of a multi-agent system, which is important while performing search, rescue, and monitoring tasks in crisis areas of various origins.

Objective is to develop a method for determining the sufficient population of a multi-agent system and the optimal routes of movement of its individual elements in a stationary network for the most complete examination of a technological disaster zone (any given zone based on a certain transport network).

Method. We implemented the concept of a dynamic programming to search for all possible edge-simple longest paths connecting the directed subsets of vertices-sources and vertices-sinks in the structure of the model weighted directed graph. To this end, the modified Dijkstra method was applied. The modification comprises representing the weights of the arcs of the modeling directed graph with the negative values, which are further used in calculations according to the Dijkstra method. After finding the next edgesimple longest path, the arcs that make up it are fixed in the memory of the computer system (in the route plan) and removed from the graph structure, and the process is iteratively repeated. The search for paths takes place as long as the transitive closure between the vertices that are part of the specified subsets of source vertices and sink vertices is preserved. The developed method makes it possible to find such a set of traffic routes for the elements of the multi-agent system, which maximizes the area examined by them in a technological disaster zone (or the number of checked objects on the traffic routes) in one “wave” of the search and distributes the elements of a multi-agent system by routes that do not have common areas. A derivative of the application of the developed method is the determination of a sufficient population of a multi-agent system for effective search activities within the defined zone.

Results. 1) A method of routing a group of mobile robots in a stationary network for searching the missing objects in a technological disaster zone has been developed. 2) The working expression of the Dijkstra method for searching in the structure of a network object (in the structure of a model graph) for the longest paths has been formalized. 3) We have suggested a set of indicators for a comprehensive evaluation of route plans of a multi-agent system. 4) The method has been verified on test problems.

Conclusions. Theoretical studies and several experiments confirm the efficiency of the developed method. The solutions made using the developed method are accurate, which allows recommending it for practical use in determining in an automated mode route plans for multi-agent systems, as well as the required number of agents in such systems to perform the required amount of search tasks in a particular crisis area.

ANALYSIS OF RISK TERMINAL FLOWS IN TECHNOGENIC SYSTEMS ARISING IN THE PROCESS OF THREAT IMPACT

2023-02-28T11:45:27+02:00

Context. The analysis of the risk terminal flows in technogenic systems is carried out, which arise in the process of the impact of informational and cognitive threats in the automated document management system as part of the hierarchical production system.

The object of the research is the process of functioning of complex systems with a hierarchical structure, in which automated document management systems with a high level of data flow protection for decision-making are used to provide the information quality control of technological processes.

The subjects of the research are the methods and means of constructing an information protection system to ensure the reliable functioning of automated document management systems and making targeted decisions in hierarchical structures with minimal risk of exposure to external threats and attacks.

Objective is to develop a complex model for assessing the risk of the document management system failure as part of a hierarchical production system under the active threats.

Method. For the first time, the cause-and-effect diagram of the event formation with the active action of threat factors and attacks is substantiated and developed, the interpretation of risk in a technogenic system is defined, and the risk in the space of states is presented as a change in the trajectory in the system transitions to the limit operation mode. For the first time, a category diagram of the structure of risk generation under the threat factors and a system-category diagram of interaction in the system risk ↔ emergencyactive nature is constructed, a system-category scheme of risk formation under the active threat factors is suggested. For the first time, a cognitive diagram for assessing losses in the event of a risk situation arising from incorrect actions of the personnel is substantiated.

Results. As a result of the research, a system-category diagram of the impact of a set of threats on the system functioning mode and process is constructed, a method is developed for calculating the level of system strategic security of energy-active hierarchical systems in the process of attacks and threats, and a complex model for assessing the risk of a system functioning failure under active threats is suggested.

Conclusions. Under the action of active obstacles, cognitive and system factors at the operational and strategic levels of the control hierarchy, due to wrong decisions and informational disorientation, emergency situations and risks of the system function loss and its target-orientation arise. The analysis of a set of risks and the suggested category diagram of the risk generation structure under the impact of threat factors form the basis of the development of the probability structure of the risk concept based on the attack ↔ consequence model, as well as the construction of a system-category diagram of the interaction in the game active factor ↔ accident risk. This, in turn, makes it possible to construct a system-category scheme for the formation of risk terminal flows in technogenic systems that arise in the process of threat impact. A complex model for assessing the risk of system failure under threats can be used to construct protection systems for any hierarchical control structures of technogenic systems.

COMPUTATIONAL INTELLIGENCE METHODS TO PATIENTS STRATIFICATION IN THE MEDICAL MONITORING SYSTEMS

2023-02-24T15:09:38+02:00

Context. In modern medical practice the automation and information technologies are increasingly being implemented for diagnosing diseases, monitoring the condition of patients, determining the treatment program, etc. Therefore, the development of new and improvement of existing methods of the patient stratification in the medical monitoring systems is timely and necessary.

Objective. The goal of intelligent diagnostics of patient’s state in the medical monitoring systems – reducing the likelihood of adverse states based on the choice of an individual treatment program:

− reducing the probability of incorrectly determining the state of the patients when monitoring patients;

− obtaining stable effective estimates of unknown values of treatment actions for patients (corresponding to the found state);

− the choice of a rational individual treatment program for the patients, identified on the basis of the forecasted state.

Method. Proposed methodology, which includes the following computational intelligence methods to patient’s stratification in the medical monitoring systems:

1) method of cluster analysis based on the agent-based approach – the determination of the possible number of patient’s states using controlled variables of state;

2) method of robust metamodels development by means artificial neuron networks under a priori data uncertainty (only accuracy of measurements is known) in the monitoring data: a) a multidimensional logistic regression model in the form of analytical dependences of the posterior probabilities of different states of the patients on the control and controlled variables of state; b) a multidimensional diagnostic model in the form of analytical dependences of the objective functions (quality criteria of the patient’s state) on the control and controlled variables of state;

3) method of estimating informativeness controlled variables of state at a priori data uncertainty;

4) method of robust multidimensional models development for the patient’s state control under a priori data uncertainty in the monitoring data in the form of analytical dependencies predicted from the measured values of the control and controlled variables of state in the monitoring process;

5) method of reducing the controlled state variables space dimension based on the analysis of the variables informativeness of the robust multidimensional models for the patient’s state control;

6) method of patient’s states determination based on the classification problem solution with the values of the control and forecasted controlled variables of state with using the probabilistic neural networks;

7) method of synthesis the rational individual patient’s treatment program in the medical monitoring system, for the state identified on the basis of the forecast.

Proposed the structure of the model for choosing the rational individual patient’s treatment program based on IT Data Stream Mining, which implements the «Big Data for Better Outcomes» concept.

Results. The developed advanced computational intelligence methods for forecast states were used in choosing the tactics of treating patients, to forecast treatment complications and assess the patient’s curability before and during special treatment.

Conclusions. Experience in the implementation of “Big Data for Better Outcomes” concept for the solution of the problem of computational models for new patient stratification strategies is presented. Advanced methodology, computational methods for a patient stratification in the medical monitoring systems and applied information technology realizing them have been developed. The developed methods for forecast states can be used in choosing the tactics of treating patients, to forecast treatment complications and assess the patient’s curability before and during special treatment.

METHOD AND SOFTWARE COMPONENT MODEL FOR SKIN DISEASE DIAGNOSIS

2023-02-24T23:56:11+02:00

Context. The problem of skin disease diagnosis was investigated in the paper. Its actuality is caused by the necessity of automation of at least advisory medical decision making. Such decisions are made in telemedicine, for instance, when skin disease diagnostics is performed under specific conditions. These conditions are specified by situations when data for analysis are collected but a qualified doctor has no possibility to process the data and to make a diagnosis decision based on it. The object of the study is a process of skin disease diagnosis.

Objective. The objective of the study is to develop a skin disease diagnosis method to automate making of advisory medical diagnosis decisions and to increase efficiency of such decisions.

Method. The skin disease diagnosis method was proposed in the work. This method applies the modified ResNet50 model. It was proposed to add layers to the ResNet50 model and to train it using transfer learning and fine-tuning techniques. The method also defines image processing in particular through the change of its resolution and uses oversampling technique to prepare a dataset for model training.

Results. Experimental investigation of the proposed method was performed using the HAM10000 dataset which contains images of skin diseases. The images were collected using dermatoscopy method. The dataset contains observations for 7 different skin diseases. The proposed method is characterized by the accuracy of 96.31% on this dataset. It is improved accuracy in comparison with the existing neural network models. Software component model was created to give a possibility to integrate the proposed method into a medical diagnosis system.

Conclusions. The obtained results of the investigation suggest application of the proposed skin disease method in medical diagnostic system to make advisory decisions by the system and to support making final decisions by a doctor.

IMAGE SEGMENTATION WITH A CONVOLUTIONAL NEURAL NETWORK WITHOUT POOLING LAYERS IN DERMATOLOGICAL DISEASE DIAGNOSTICS SYSTEMS

2023-02-25T15:08:29+02:00

Context. The problem of automating of the segmentation of spectral-statistical texture images is considered. The object of research is image processing in dermatological disease diagnostic systems.

Objective. The aim of the research is to improve the segmentation performance of color images of psoriasis lesions by elaboration of a deep learning convolutional neural network without pooling layers.

Method. The convolutional neural network is proposed to process a three-channel psoriasis image with a specified size. The initial color images were scaled to the specified size and then inputed on the neural network. The architecture of the proposed neural network consists of four convolutional layers with batch normalization layers and ReLU activation function. Feature maps from the output of these layers were inputted to the 1*1 convolutional layer with the Softmax activation function. The resulting feature maps were inputted to the image pixel classification layer. When segmenting images, convolutional and pooling layers extract the features of image fragments, and fully connected layers classify the resulting feature vectors, forming a partition of the image into homogeneous segments. The segmentation features are evaluated as a result of network training using ground-truth images which segmented by an expert. Such features are robust to noise and distortion in images. The combination of segmentation results at different scales is determined by the network architecture. Pooling layers were not included in the architecture of the proposed convolutional neural network since they reduce the size of feature maps compared to the size of the original image and can decrease the segmentation performance of small psoriasis lesions and psoriasis lesions of complex shape.

Results. The proposed convolutional neural network has been implemented in software and researched for solving the problem of psoriasis images segmentation.

Conclusions. The use of the proposed convolutional neural network made it possible to enhance the segmentation performance of plaque and guttate psoriasis images, especially at the edges of the lesions. Prospects for further research are to study the performance of the proposed CNN then abrupt changes in color and illumination, blurring, as well as the complex background areas are present on dermatological images, for example, containing clothes or fragments of the interior. It is advisable to use the proposed CNN in other problems of color image processing to segment statistical or spectral-statistical texture regions on a uniform or textured background.

MACHINE LEARNING DECISION SUPPORT SYSTEMS FOR ADAPTATION OF EDUCATIONAL CONTENT TO THE LABOR MARKET REQUIREMENTS

2023-02-25T19:16:12+02:00

Context. The urgent task of increasing the functional efficiency of machine learning of decision support system (DSS) for assessing compliance with content modern requirements of the educational disciplines of the graduation department based on the results of the employer survey has been solved.

Objective. Increasing the functional efficiency of machine learning of DSS for assessing compliance with modern requirements of the educational disciplines content of the first (bachelor’s) level specialty educational and professional program based on machine learning and pattern recognition.

Method. The method of machine learning of DSS is proposed for adapting the educational content of the graduation department to the labor market requirements. The idea of the method is to maximize the information capacity of the DSS in the machine learning process, which allows in the monitoring mode to guarantee a high full probability of making the correct classification decisions. The method was developed as part of a functional approach to modeling cognitive processes of natural intelligence, which makes it possible to provide DSS with flexibility when retraining the system due to increasing the power of the recognition classes alphabet. The method is based on the principle of maximizing the amount of information in the machine learning process. The modified Kullback information measure, which is a functional of the accuracy characteristics of classification solutions, is considered as a criterion for optimizing machine learning parameters. According to the proposed functional category model, an information-extreme machine learning algorithm was developed based on the hierarchical data structure in the form of a binary decursive tree. The use of such a data structure allows you to automatically divide a large number of recognition classes into pairs of nearest neighbors, for which optimization of machine learning parameters is carried out according to a linear algorithm of the required depth. The geometric parameters of hyperspherical containers of recognition classes were considered as optimization parameters, which were restored in the radial basis of the binary space of Hamming features in the machine learning process. At the same time, the input traning matrix was transformed into a working binary training matrix, which was changed in the machine learning process through admissible transformations in order to adapt the input information description of the DSS to the maximum reliability of classification decisions.

Results. The informational, algorithmic, and software of the DSS was developed to assess the educational content quality based on the machine analysis results of respondents’ answers. Within the framework of the geometric approach, based on the informationextreme machine learning results, highly reliable decisive rules, practically invariant to the multidimensionality of the recognition features space, were constructed based on the hierarchical data structure in the form of a binary decursive tree. The influence of machine learning parameters on the functional effectiveness of machine learning of the DSS was studied on the evaluation example of the educational content of the educational and professional bachelor’s program of the specialty 122 Computer Science.

Conclusions. The computer modeling results confirm the high functional efficiency of the proposed method of informationextreme hierarchical machine learning and can be recommended for practical use in institutions of higher education to assess compliance with modern requirements of the educational content of graduation departments.

CLUSTERIZATION OF DATA ARRAYS BASED ON THE MODIFIED GRAY WOLF ALGORITHM

2023-02-25T19:42:36+02:00

Context. The task of clustering arrays of multidimensional data, the main goal of which is to find classes of observations that are homogeneous in the sense of the accepted metric, is an important part of the intelligent data analysis of Data Mining. From a computational point of view, the problem of clustering turns into the problem of finding local extrema of a multiextreme function, which are repeatedly started from different points of the original data array. To speed up the process of finding these extrema using the ideas of evolutionary optimization, which includes algorithms inspired by nature, swarm algorithms, population algorithms, etc.

Objective. The purpose of the work is to introduce a procedure for clustering data arrays based on the improved gray wolf algorithm.

Method. A method of clustering data arrays based on the modified gray wolf algorithm is introduced. The advantage of the proposed approach is a reduction in the time of solving optimization problems in conditions where clusters are overlap. A feature of the proposed method is computational simplicity and high speed, due to the fact that the entire array is processed only once, that is, eliminates the need for multi-era self-learning, implemented in traditional fuzzy clustering algorithms.

Results. The results of the experiments confirm the effectiveness of the proposed approach in clustering problems under the condition of classes that overlap and allow us to recommend the proposed method for use in practice to solve problems of automatic clustering big data.

Conclusions. A method of clustering data arrays based on the modified gray wolf algorithm is introduced. The advantage of the proposed approach is the reduction of time for solving optimization problems. The results of the experiments confirm the effectiveness of the proposed approach in clustering problems under the conditions of overlapping clusters.

METHOD OF SELF-DEFENSE OF GROUND (SURFACE) OBJECTS FROM HIGH-PRECISION RADAR MEANS OF AIR SURVEILLANCE

2023-02-24T12:57:21+02:00

Context it is caused by the need to search for scientific and technical ways to ensure the effectiveness of protecting ground (surface) objects from high-precision guided missile weapons.

Objective it is a necessity to ensure effective self-defense of objects from radar homing means.

Method. Electrodynamic modeling of Echo signals from spatially distributed objects, taking into account the features of their design and related operational limitations.

Results. Based on the analysis of the shortcomings of the well-known method of protecting stationary objects from radar surveillance and damage, based on the simulation of an effective reflection center outside the physical dimensions of the object, a new method of countering high-precision measurement of coordinates of stationary and mobile ground (surface) objects is proposed. The technique is based on the spatial deformation of the location of the effective target reflection center with dynamics that exceed the inertial capabilities of the auto-observation contour of the attacking missile (projectile). A structural and functional scheme of technical implementation of the methodology based on the first proposed relationship of simple design and technological solutions is proposed and justified.

Conclusions. The analytical model of Echo signals of spatially distributed ground (surface) objects was further developed, which takes into account the specifics of their design, and on its basis, for the first time, a universal method of self-defense of objects from radar home-leading devices was developed, which is implemented in a patented method and complex to exclude damage to protected objects.

PARAMETER-DRIVEN GENERATION OF EVALUATION PROGRAM FOR A NEUROEVOLUTION ALGORITHM ON A BINARY MULTIPLEXER EXAMPLE

2023-02-26T18:18:11+02:00

Context. The problem of automated development of evaluation programs for the neuroevolution of augmenting topologies. Neuroevolution algorithms apply mechanisms of mutation, recombination, and selection to find neural networks with behavior that satisfies the conditions of a certain formally defined problem. An example of such a problem is finding a neural network that implements a certain digital logic.

Objective. The goal of the work is the automated design and generation of an evaluation program for a sample neuroevolution problem (binary multiplexer).

Method. The methods and tools of Glushkov’s algebra of algorithms and hyperscheme algebra are applied for the parameterdriven generation of a neuroevolution evaluation program for a binary multiplexer. Glushkov’s algebra is the basis of the algorithmic language intended for multilevel structural design and documentation of sequential and parallel algorithms and programs in a form close to a natural language. Hyperschemes are high-level parameterized specifications intended for solving a certain class of problems. Setting parameter values and subsequent interpretation of hyperschemes allows obtaining algorithms adapted to specific conditions of their use.

Results. The facilities of hyperschemes were implemented in the developed integrated toolkit for the automated design and synthesis of programs. Based on algorithm schemes, the system generates programs in a target programming language. The advantage of the system is the possibility of describing algorithm schemes in a natural-linguistic form. An experiment was conducted consisting in the execution of the generated program for the problem of evaluating a binary multiplexer on a distributed cloud platform. The multiplexer example is included in SharpNEAT, an open-source framework that implements the genetic neuroevolution algorithm NEAT for the .NET platform. The parallel distributed implementation of the SharpNEAT was proposed in the previous work of the authors.

Conclusions. The conducted experiments demonstrated the possibility of the developed distributed system to perform evaluations on 64 cloud clients-executors and obtain an increase in 60–100% of the maximum capabilities of a single-processor local implementation.

THE CURVE ARC AS A STRUCTURE ELEMENT OF AN OBJECT CONTOUR IN THE IMAGE TO BE RECOGNIZED

2023-02-26T18:43:10+02:00

Context. The proposed article relates to the field of visual information processing in a computer environment, more precisely to the determination the parameters of the interest object in the image, in particular, the contour of the interest object In most cases, the contour of an object is a simply connected sequence of curve arcs.

Objective. The purpose and subject of the study is to find and to propose such a definition of the digital curve arc, as the most important element of the object contour in the recognizable image, which does not contradict modern neurophysiological conceptions about visual perception, and to recognize the object contour as a sequence of the digital curve arcs.

Method. The representation of the image in the form of a structural model is used, one of the structural elements of which is the contour of the object, consisting of digital curve arcs. Also, the image is considered as a cellular complex which corresponds to modern ideas about human visual perception.

Results. The new definition for arc of a digital curve as a sequence of digital straight segments is proposed, which does not contradict to modern concepts of neurophysiology. In contrast to the known definitions of a curve arc, the proposed definition of a digital curve arc makes it possible to determine the start and end points of the arc. According to the description of the contour of an object as a simply connected closed sequence of line segments, it is proposed to construct a description of the contour as a sequence of arcs of digital curves.

Conclusions. The use of the proposed definition of the digital curve arc in image processing makes it possible to recognize the contour of an object in an image and present it in a form close to visual perception. For best results, the use of variable resolution in image processing algorithms is recommended.

APPLICATION OF TWO-DIMENSIONAL PADÉ-TYPE APPROXIMATIONS FOR IMAGE PROCESSING

2023-02-26T21:24:07+02:00

Context. The Gibbs phenomenon introduces significant distortions for most popular 2D graphics standards because they use a finite sum of harmonics when image processing by expansion of the signal into a two-dimensional Fourier series is used in order to reduce the size of the graphical file. Thus, the reduction of this phenomenon is a very important problem.

Objective. The aim of the current work is the application of two-dimensional Padé-type approximations with the aim of elimination of the Gibbs phenomenon in image processing and reduction of the size of the resulting image file.

Method. We use the two-dimensional Padé-type approximants method which we have developed earlier to reduce the Gibbs phenomenon for the harmonic two-dimensional Fourier series. A definition of a Padé-type functional is proposed. For this purpose, we use the generalized two-dimensional Padé approximation proposed by Chisholm when the range of the frequency values on the integer grid is selected according to the Vavilov method. The proposed scheme makes it possible to determine a set of series coefficients necessary and sufficient for construction of a Padé-type approximation with a given structure of the numerator and denominator. We consider some examples of Padé approximants application to simple discontinuous template functions for both formulaic and discrete representation.

Results. The study gives us an opportunity to make some conclusions about practical usage of the Padé-type approximation and about its advantages. They demonstrate effective elimination of distortions inherent to Gibbs phenomena for the Padé-type approximant. It is well seen that Padé-type approximant is significantly more visually appropriate than Fourier one. Application of the Padétype approximation also leads to sufficient decrease of approximants’ parameter number without the loss of precision.

Conclusions. The applicability of the technique and the possibility of its application to improve the accuracy of calculations are demonstrated. The study gives us an opportunity to make conclusions about the advantages of the Padé-type approximation practical usage.

THE METHOD OF ASSESSING THE VALUE OF INFORMATION

2023-02-27T11:44:20+02:00

Context. The task of assessing the value of the institution’s information as one of the objects of protection of the information security model is considered.

Objective. The goal of the work is the creation of a method of assessing the value of information, which takes into account the time of the final aging of information.

Method. The results of the analysis of methods for evaluating the value of information showed that modern approaches are conventionally divided into two directions. In the first direction, the value of information is calculated as the amount of information in bytes. In the second direction, the value of information is calculated in monetary terms. It is shown that modern approaches do not take into account the influence of time on the value of information. A method of assessing the value of information is proposed, which takes into account such characteristics as the term of final aging of information, the level of its access restriction, importance, and form of ownership. The value of information is presented as a quantitative measure that determines the degree of its usefulness for the owner. It is proposed to calculate the value of the initial value of information during its creation or acquisition by calculating the normalized weight of the coefficients according to the formula of the arithmetic mean. It was shown that the current value of information has a functional dependence on the time of existence of information and the time of its final aging.

Results. The results of the experiment confirm that the value of information has a nonlinear functional dependence on the time of final aging of information.

Conclusions. The conducted experiments confirmed the efficiency of the proposed method of evaluating the value of information and allow recommending it for use in practice to protect the institution’s information. Prospects for further research may include the creation of a methodology for assessing the value of an institution’s information, taking into account the aging of information and subsequent adjustment of measures to protect it.

TECHNOLOGY FOR GRAMMATICAL ERRORS CORRECTION IN UKRAINIAN TEXT CONTENT BASED ON MACHINE LEARNING METHODS

2023-02-27T12:15:51+02:00

Context. Most research in grammatical and stylistic error correction focuses on error correction in English-language textual content. Thanks to the availability of large data sets, a significant increase in the accuracy of English grammar correction has been achieved. Unfortunately, there are few studies on other languages. Systems for the English language are constantly developing and currently actively use machine learning methods: classification (sequence tagging) and machine translation. A large amount of parallel or manually labelled data is required to build a high-quality machine learning model for correcting grammatical/stylistic errors in the texts of those morphologically complex languages. Manual data annotation requires a lot of effort by professional linguists, which makes the creation of text corpora, especially in morphologically rich languages, mainly Ukrainian, a time- and resource-consuming process.

Objective of the study is to develop a technology for correcting errors in Ukrainian-language texts based on machine learning methods using a small set of annotated parallel data.

Method. For this study, machine learning algorithms were selected when developing a system for correcting errors in Ukrainianlanguage texts using an optimal pipeline, including pre-processing and selecting text content and generating features in small annotated data corpora. The neural network’s use with a new architecture, a review of state-of-the-art methods, and a comparison of different pipeline stages will make it possible to determine such a combination of them, allowing a high-quality error correction model in Ukrainian-language texts.

Results. A machine learning model for error correction in Ukrainian-language texts has been developed. A universal scheme for creating an error correction system for different languages is proposed. According to the results, the neural network can correct simple sentences written in Ukrainian. However, creating a full-fledged system will require spell-checking using dictionaries and checking rules, both simple and based on the result of parsing dependencies or other features. The pre-trained neural translation model mT5 has the best performance among the three models. To save computing resources, it is also possible to use a pre-trained BERT-type neural network as an encoder and a decoder. Such a neural network has half the number of parameters as other pretrained machine translation models and shows satisfactory results in correcting grammatical and stylistic errors.

Conclusions. The created model shows excellent classification results on test data. The calculated machine translation quality metrics allow only a partial comparison of the models since most of the words and phrases in the original and corrected sentences are the same. The best value for both BLEU (0.908) and METEOR (0.956) is obtained for mT5, which is consistent with the case study in which the most accurate error corrections without changing the initial value of the sentence are obtained for such a neural network. The M2M100 has a higher BLEU score (0.847) than the “Ukrainian Roberta” Encoder-Decoder (0.697). However, subjectively evaluating the results of the correction of examples, the M2M100 does a much worse job than the other two models. For METEOR, M2M100 (0.925) also has a higher score than the “Ukrainian Roberta” Encoder-Decoder (0.876).