Search

Predictive psychodiagnostics is the process of forming a predictive assessment of any metric based on the data of psychodiagnostic testing. The formation of a predictive estimate is a classification problem and a variety of algorithmic approaches can be used to solve it – logistic regression, support vector machines, decision trees, neural networks, and many others. The article describes the practice of using algorithms based on decision trees to create models used in solving predictive psychodiagnostics problems, as well as cases in which their application is most relevant. The results obtained using the models built on the basis of the decision tree algorithm using gradient boosting show sufficient predictability and statistical significance in comparison with the methods of traditional psychodiagnostics used to solve personnel assessment problems

The article deals with the architecture and structure of the most effective artificial neural networks (ANNs) in the tasks of network traffic classification: multilayer perceptron and LSTM (long short-term memory) network. Using one of the actual datasets CSE-CIC-IDS2018 trained ANNs and tested the effectiveness of detection of various attacks in the network traffic. An analysis of existing work has been performed. A multiclass classification was performed and the detection effectiveness of separate computer attacks was reported. The effectiveness evaluation focused on three metrics: completeness, correct response rate, and false positive rate. The test classification results were considered both in terms of comparison of numerical characteristics of recognition quality and in terms of class analysis in order to identify which computer attacks (CA) are better identified using multilayer perceptron and which are better identified using LSTM. On the basis of these results, an ANN, which is a hybrid network, is proposed as a promising Intrusion Detection System (IDS).

Currently, unmanned vehicles cannot yet be introduced on public roads, and the person behind the wheel is not predictable and behavior is a multi-parameter task. The article describes a methodology for calculating and evaluating various characteristics of a respondent's psychodiagnostic test using the integrated use of typical driver behavior uncertainty and assessing his condition through machine learning (ML) algorithms. The application of machine learning allows you to reduce the number of test questions and increase the processing speed, since the developed algorithm finds hidden patterns well. A large number of factors affect the driver's condition, and this article is devoted to determining the driver's condition in real time, which reduces the likelihood of an emergency. Based on the developed algorithm, it is possible to conduct an interview and pre-trip control. On the one hand, this makes it possible to improve the quality of the recruited personnel, and on the other, to improve the quality of logistics and, most importantly, to increase road safety.

The article describes an approach to reducing the number of questions in the Keys to Personal Mastery Inventory (KPMI) psychological testing questionnaire with the further addition of missing data and their use in a machine learning model to predict respondent satisfaction with their job. Since the 30s of the 20th century, mathematicians, physicists and engineers have been working on the description of traffic flows. But theoretical models do not work well because a person is driving a car, and a person is unpredictable. In the same situation, it behaves differently. This work aims to reduce driver uncertainty using machine learning methods.

Information security systems that use machine learning algorithms and are based on real datasets are among the most effective. The format and content of datasets used in artificial intelligence and machine learning methods directly affect the quality of information security system implementation. The article presents a classification of datasets in the field of information security of communication networks. The structure and features of several characteristic data sets related to the use of machine learning algorithms in information security systems are shown.

Ensembles of neural networks are used to improve the performance of the model and show themselves well in the case of using deep recurrent and convolutional networks on serial [1] and graphic data [2]. The article discusses the advisability of using a snapshot ensemble of not deep feedforward neural networks on a synthetic dataset.

As part of the study of mobile device traffic, the formation of an experimental database of network traffic of selected mobile applications, which can be used for training and testing classifiers using machine learning methods, occupies an important place.To automate the process of studying mobile application traffic classification algorithms, a software package has been developed that allows you to automatically collect network traffic packets from mobile devices and save them to a database; group network traffic packets into streams; upon user request, generate data sets with specified characteristics (number of specific application flows: background traffic; generate a data set based on the already created set with the addition of new flows, excluding repetitions).To collect traffic from mobile devices running the Android operating system, an application was developed that uses the application programming interface to create virtual private networks, collects network traffic packets, identifies the source application and sends them via HTTP to server software.Using the client and server software, the database created was filled with traffic from 18 main applications. During the data collection, 71,667 streams and 6,989,991 packets were received.

One of the tasks of system identification within the modern signal processing system is the task of identification of unknown error-correction code (ECC). System identification can be done using various methods. Anyway, the more knowledge one has about the possible system structure provides better solution for system identification. Here we take a look into digital communication systems uses ECC, which are the major part of modern communication systems are. We create a new practical approach for identification of a priori unknown ECC being used over the transmission channel of the communication system which is examined for identification purposes. The approach is based as Big Data application which can provide new source for the study. Shown that usage of new approach provides better speed for system identification making the process of identification faster, more reliable, with better quality for solution found and with lower requirements for quality of the transmission channel.

A model is proposed which describes the electrostriction temporal shift of laser pulses in their high-frequency train propagating in an optical fibre. The validity of the model is confirmed by good agreement of its predictions with the results of studies of relative temporal shifts in the high-frequency train of optical solitons known from the literature. The model predicts the possibility of a substantial reduction of temporal shifts in the data pulse train in fibreoptic links composed of fibres with different signs of dispersion. For comparatively low pulse repetition rates (less than 500 MHz), the oscillating dependence of the relative temporal shift of pulses on the data transmission rate is found.