CC BY
19Список литературы
1. Официальный сайт Роспотребнадзора республики Мордовия «Управление федеральной службы по надзору в сфере защиты прав потребителей и благополучия человека по республике Мордовия» [Электронный ресурс]. - Загл. с экрана. - режим доступа: http://13.rospotrebnadzor.ru/center/ser-vices/zdorov_obraz/135871.
2. Сайт торговой сети «Ситилинк» [Электронный ресурс]. - Загл. с экрана. - режим доступа:
https://www.citilink.ru/catalog/computers_and_note-books/computers/1162618/certificates/.
3. Рысин, Ю. С. Безопасность жизнедеятельности. Электромагнитное излучение: учебное пособие / Ю. С. Рысин, А. К. Сланов, С. Л. Яблочников.
— Саратов: Ай Пи Эр Медиа, 2019. — 82 c. — ISBN 978-5-4486-0584-0. — Текст: электронный // Электронно-библиотечная система IPR BOOKS: [сайт].
— URL: http://www.iprbookshop.ru/80169.html.
METHODS AND MODELS FOR PASSENGER TRANSPORTATIONS FORECASTING ON AIR
ROUTES
Volkovska A.
senior lecturer, Air Transportation Management Department Faculty of Transport, Management and Logistics of the National Aviation University
Abstract
Article considers basic methods and models implemented in forecasting air passenger transportations. Main influencing factors, advantages and disadvantages of each method are identified.
Keywords: air passenger transportation, forecast, gravitational model, regression and correlation analysis, statistical forecasting of time series, heuristic forecasting.
At the end of 19th century, A.Wellington (USA) and E.Lille (Austria-Hungary) first proposed imlemen-tation of gravitational models to forecast passenger flows. Lille found that frequency of travel varies inversely with the square of distance traveled, and Wellington calculated that passenger flow between cities is equal to the square of their population divided by the distance between points. Each of researchers (on example of several cities) with some error found confirmation of their empirical dependencies, introducing additional coefficients. But to extend these models as common to forecast transportation volumes between correspondent points proved impossible.
Researchers M.I.Zagordan and F.P.Kravets in 1932 proposed for forecasting model in which passenger flow was as product fraction of population in corresponding cities by square of distance between them. The authors tried to take into account influence of a number of factors on passenger flow by multiplying it by correction factor, which differs significantly for different cities (3-5 times).
In later foreign studies, gravitational model was
used
kHH
A, = —— (1)
V ra v '
ll
where A - volume of passenger flow on the line
between points i andj ; k is empirical coefficient; Hi, Hj - population in cities i and j ; Lj - distance between cities i and j ; a is a factor that reflects the dependence of number of trips on distance.
Gravitational models allow to achieve a certain level of accuracy only in specific conditions and for a short period of time. Although relative simplicity of obtaining initial data on population and distance is posi-
tive feature of such models, but difficulty of determining correction factor, which must take into account many factors and individual characteristics of corresponding points, does not allow extensive use of gravity models in forecasting. If transportation between points is performed by several transport modes, then there is a problem of passenger flow distribution between them, which further complicates implementation of gravitational models.
Further research in the field of passenger flows forecasting was aimed at identifying factors influencing the demand of population in various types of connections. Having data on future quantitative changes in some major factors and knowing how they affect passenger flow, it is possible to predict its value.
Elasticity of air transportations volumes
dA k,dF,
T = (2)
A F
where A is air transportations volume; dA - increase in air transportations; k - coefficient of elasticity of air transportations from i -th factor ; Fi - quantitative characteristics of i -th factor ; dFi is increase in factor feature.
Coefficient of elasticity shows change ratio in percentage of one feature (air transportations) with 1% change of another feature (income). Coefficient of elasticity can determine possible change in transportations volume in calculation period in accordance with income dynamics.
Coefficient of elasticity changes over time, so it is necessary to use this model carefully in long-term forecasting, especially when changing economic conditions (for example, tariffs). Determining coefficient of elasticity is associated with large amount of statistical information processing, which, as a rule, does not reflect real demand for air transportation. Therefore, results of
forecast will have an error, degree of which is difficult to quantify due to the lack of data on unmet demand.
In modern conditions, when substantiating volume of passenger transportations, two types of forecasts are used: forecast of demand for air transportation and forecast of possible volumes of transportations (passenger turnover), obtained on the basis of data on actual demand. The difference between these values is amount of unmet demand. In practice, air transportations forecasts in most cases focus on data on actual traffic. Further improvement of passenger transportations forecasting methods is aimed at identifying unmet demand, and planning - to meet it in economically justified amounts.
Since the main task of planning in modern conditions is to focus on final needs, justification of air transportations should be based on data on the total demand for all transport modes with subsequent establishment of air transport share in the development of total passenger transportations. Taking into account this requirement and methods of forecasting passenger transportations should be developed.
Scientific forecasting has a large number of methods and techniques of forecasting, of which the most widespread in air transport are methods: mathematical modeling based on regression and correlation analysis; statistical forecasting of time series (extrapolation); heuristic forecasting (expert assessments).
The first group combines methods of regression and correlation analysis, which consider development of forecast process depending on influence of certain factors. In these analytical models, time is not usually used as a factor.
The second group includes methods based on extrapolation of time series using various analytical functions that characterize the process sequentially over time.
Heuristic forecasting methods are based on processing of expert assessments of economic process expected development. They can be used in conjunction with the first two methods.
Depending on the term of forecast period, following types of air transportations forecasts are distinguished: long-term (for 10 years or more); medium-term (up to 5 years); short-term (per year, quarter, month, decade, day).
Air transportations forecasting system is tool for implementing given planning mode. Forecast decisions (long-term, medium-term, current and operational) enter the planning system, where they are based on planned indicators calculation.
For long-term planning, air transportations volumes should be forecasted at the end of each five-year period and over longer period. When developing five-year forecast, forecasting is carried out for each year of planning period.
Current planning provides for development of annual transportations volumes forecast broken down by quarters of the year. Operational forecasting provides initial data for planning for shorter periods of time (quarter with breakdown by months, decade, day).
Implementation of different forecasting methods to substantiate of transportations volume depends on
availability of initial data (information support), study degree of factors influencing passenger flows formation, mathematical support of tasks (models, programs), level of methodological developments. The main methods of air transportations forecasting are discussed below.
Forecasting based on methods of correlation and regression analysis is reduced to finding regression equation (paired or multiple), in which transportations volume or passenger turnover (dependent variable) is depending on one or group of factors. Regression model is built by processing reporting data on transportation (passenger turnover), presented in the form of time series and corresponding values of passenger-forming factors (population, its monetary income, level of tariffs, etc.).
In general case, transportations volume or passenger turnover can be expressed depending on a number of factors as follows:
An =f(xi, X2, x3,..., Xn) (3)
where An - transportations volume (or demand for
transportation);
X\ , Xf}
- factors affecting air transportation volume.
Regression analysis involves solving the following tasks: analysis and selection of factors that have a significant impact on dependent variable; choice of the form of regression equation; finding the empirical parameters of equation; statistical evaluation and adjustment of model parameters.
Construction of mathematical model is preceded by logical analysis and processing of statistical data. Uncharacteristic values of indicators should not always be excluded from the time series, as they can be accepted as threshold (or turning points) in the transportations development.
After the most significant factors have been identified by means of qualitative analysis, it is necessary to find out whether there are any linearly dependent ones among them, i.e. to check factors for multicollinearity. In the presence of such relationships, estimates of regression equation parameters lose stability. It is believed that there is significant multicollinearity between factor features if correlation coefficient between them reaches 0.8. To establish the degree of relationship significance between factors, matrix of correlation coefficients is calculated, in analysis of which the presence of multicollinearity is determined. In some models proposed for forecasting passenger transportations, as independent variables take simultaneously the gross domestic product, fixed assets of economy, real incomes, time. It is easy to establish that there is an interdependence between individual factors in such models. For example, increase in value of fixed assets leads to increase in gross domestic product, which, in turn, is accompanied by increase in real incomes. Finally, dynamics of all these indicators can be expressed as a function of time.
One of the requirements for constructing regression models is that the type of dependence be as simple as possible, because complex functions create additional difficulties in calculating parameters of model
and its evaluation. That is why when forecasting transportations, one should strive to choose the most significant factors.
Currently, significant number of models have been developed that are used to forecast the passenger transportations volume. All of them can be combined into two large groups:
- general (undirected) models for forecasting transportations volume and passenger turnover in the country as a whole, economic regions, territorial administrations, enterprises;
- partial (directed) models of passenger flows forecasting by specific correspondence (connections) between cities (settlements), airlines, groups of directions (airlines).
The basis of passenger transportations calculation and passenger turnover in some models is population mobility indicator which is defined depending on passenger-forming factors.
Passenger turnover (AL) b and population mobility Pb in air transport in general models are determined by formulas:
(AL)B = PBH; PB = a0 + + ah +£t (4)
where H is population; xi, X2 - factors (for example, real per capita income and share of urban population); a0, ai, a2 - estimation parameters of model; et -random variable that characterizes deviation from theoretical line.
It is possible to use model to forecast shipments by local airlines
y=a0 + ax + ax2 + a x + a4x4 (5)
where xi - population; X2 - gross domestic product or real income ; X3 - length of roads; X4 - length of railways; ao, ai, a2, 03, 04 - parameters of model, which are calculated by least squares method.
When selecting factors influencing dynamics of passenger transportations, it is important to take into account that reliability of forecast results does not depend on their number, but on how reasonably selected main factors that determine transportations growth. Such factors are usually few. And it should be emphasized that so far no such main indicators have been found that would directly determine traffic volume and passenger turnover. At the present stage of passenger transport development, passenger flows are determined by population, its dispersion throughout the country, need to move, availability of travel time and opportunities to pay for transport services.
Population determines the total mass of traffic, and its increase causes a constant increase in transport products. Thus, population is one of the main passenger-forming factors. In the absence of other factors influencing traffic increase, obviously, increase in population would determine entire increase in traffic.
To predict the volume of passenger transportations in specific correspondence in the presence of parallel other connections types can be used expressions:
Abi = YBPMA; ABi = PBlH (6)
where yBi - aviation ratio; PMi, PBi - population
mobility in the /-th direction, respectively, on all transport modes and air transport; H - population,
which determines passengers flow.
If aviation direction connects points of approximately identical classification (administrative-industrial centers), then population of both cities (points) is taken into account. If one of the points is resort center, departures are forecasted based on the population mobility of the resort town.
Population mobility in specific areas can be forecasted by processing traffic statistics and identifying factors that affect their dynamics.
Mobility index is usually forecasted with the increase of money incomes and costs, and aviation ratio -based on its dynamics for several years.
Traffic volume in a particular direction in the presence of statistical data on traffic can be determined by formula
ab, = X A K kn knpyBl (7)
where X As - total traffic volume by all transport modes in the base year; kH - coefficient that
takes into account impact of population growth; kn -coefficient that takes into account growth of population mobility; knp - coefficient that takes into account action of other factors; yBi - factor aviation ratio in the
reported year.
Coefficients that take into account passenger-forming factors are determined by different methods. Forecasting aviation ratio is complicated by the lack of complete statistical reporting on interstation correspondence on rail transport. Therefore, further work is needed on different types of passenger transport to process primary data on transportation and generalize them for forecasting.
In the process of forecasting air transportation by correspondence (airlines), there may be instability in the dynamics of shipments in some directions, especially on routes with small traffic volumes. Therefore, forecast is made only for the main, most powerful passenger flows, and other directions are combined, and they forecast total shipments volume, i.e.
X AB = AB1 + AB2 + ••• + ABl + X AB.other (8) where X AB - forecast of shipments through airport ; AB1, AB2,...,Ai - forecasts for the main directions; X Ab, other - total forecast of shipments in other
directions. Airlines and directions can be
grouped according to various features: distance and growth rates of traffic, specific features of airlines (resort destinations, internal communications, etc.). When grouping directions on range forecast model can be done:
X AB = X AB1 + X AB 2 + ••• + X ABi (9)
where ^AB1,^ABAB. - forecasts of transportations on 1,2..,i-th direction groups.
When forecasting may be cases when there are no air traffic statistics (new airline) and it is not possible to determine the actual aviation ratio. Then principle of similarity can be used, i.e. it is necessary to choose an airline with similar characteristics.
Time series extrapolation is the most common and most developed statistical method of forecasting. Essence of extrapolation method is that on the basis of statistical data analysis for a number of years establish patterns of indicator change and development trend, which is represented as an equation. Then determine value of indicator outside the time series. This method is based on premise that established trend will continue in the future.
Extrapolation method includes the following steps:
1. analysis and statistical processing of time series;
2. choice of dependence type that gives closest to the actual values of series;
3. determination of selected dependence parameters;
4. obtaining forecast and calculating confidence intervals.
When analyzing time series, following conditions must be taken into account:
- initial data must be qualitatively homogeneous and not have sharp jumps;
- time series should reflect sufficiently long and characteristic period of time for trend reliability;
- statistical information should not have gaps within time series.
The most important point is choice of equation that characterizes change trend of indicator under consideration. Dependence is chosen by analyzing data based on accepted evaluation criteria. Often criteria is the sum of deviation squares of indicator actual values from calculated ones obtained in accordance with equations under consideration. Dependence that has the least criteria value is selected. But such assessment does not preclude qualitative analysis of equations under consideration, in essence, their compliance with the nature of forecasted indicator. Therefore, final decision on acceptability of equation is made by the forecaster.
Another method that allows to justify the choice of dependence is to compare characteristics of change in actual data increments with corresponding characteristics of growth curves. Dependence is chosen, law of growth change of which is closest to regularity of actual data change.
Method of time series approximation by n- th degree polynomials is the most developed in theory. Equation of polynomial, which describes changes in traffic rate over time, has form:
y = a + at+at2 + •••+aJn (io)
The least squares method is used to find parameters of equation. At each increase of dependence order
not only definition of new parameter a„, but also enumeration of all other parameters as equation system changes is required.
Practically at finding of air transportations volumes the following dependences are most often used: linear: y=a+bt parabola: y=a+bt+ct2 logarithmic: y=a+blnt degree: y=atb
modified exponent: y=k-aebt Other equations can be chosen depending on the nature of time series.
Due to the fact that extrapolation in traffic volumes forecasts preserves past trends, it is advisable to consider them not as the end result, but as some initial moment, on the basis of which with involvement of additional information more reasonable solutions are developed.
Statistical forecasting methods include extrapolation according to average growth rate of transportations volume (passenger turnover). Forecasted transportations volume:
y , = y6 Kn (11)
where y6 - transportations volume in base (last) year; K - average annual growth rate; n is the number
of years of forecast period.
Average annual growth rate is calculated by formula:
K = .
m
y±
y»
(12)
where yH - transportations volume in initial (first)
year of series; m is the number of series intervals.
Heuristic forecasting methods (HFM) are of
great importance for air transportations forecasting. Their essence consists in processing of forecast estimations received from highly skilled experts in the field of air transportations organization and planning. Forecast estimates reflect individual judgments of specialists regarding development of transportation and are based on their experience and intuition.
In heuristic forecasting (expert assessments) are
used:
- methods of individual assessments;
- methods of collective assessment. Expert assessments are required if:
- there are no reporting (statistical) data, or they are insufficient;
- there are no sufficiently reliable statistical methods of assessment based on data from previous periods;
- in the development of process there are dramatic changes, characteristics of which are little known;
- it is necessary to evaluate forecasts obtained by other methods.
Positive qualities of method include its versatility: possibility of use in any time intervals and to assess development of various forecasting objects; application at difficulties with application of other methods; use as a stand-alone method and in combination with others.
Disadvantages of method: more complex than other methods, organizational support; influence of specialists-experts competence level; presence of subjective factor.
Positive properties of extrapolation method include: availability of statistical and other information on transportations volume of required detail, developed mathematical apparatus and software for forecasting operations.
Main disadvantages: transfer to the future of past trends that may not be sufficiently confirmed and lack of connection with passenger-forming factors.
Advantages of methods based on correlation and regression analysis are accounting in forecast models of passenger-forming factors, sufficient methodological support (mathematical), reliability of results with correctly selected factors, ability to use at any time forecasting.
Main disadvantage of method is difficulty of obtaining information about numerical values of passenger-forming factors and reducing reliability of results with formal approach to passenger-forming factors choice.
For air transportations planning, it is advisable to use different methods of forecasting. Preference should be given to regression models and methods of heuristic forecasting, as they take into account influence of real factors on passenger flows formation processes.
References
1. Yaschenko L.A., Shapoval N.S., Merzhvin-skaya A.N. Feasibility studies and forecasting industry development. Tutorial - K.: Center for educational literature, 2006 - 240 p.
2. Kulaev Y.F. Economic evaluation of investment projects of technical and for the air transport. Brief Guidelines - K.: KMUHA, 1996 - 16 p.
INTERNET OF THINGS (IoT) AND NEURAL NETWORKS INTERACTION DURING VIDEO
OPERATION SURVEILLANCE SYSTEMS
Krutko D.
Siberian State University of Science and Technology named after Academician M.F. Reshetnev (Krasnoyarsk), student Khodenkova E.
Siberian State University of Science and Technology named after Academician M.F. Reshetnev (Krasnoyarsk), Candidate of Philosophy
Abstract
Our future is progressing more and more rapidly under the influence of such trends as the Internet of Things, neural networks and hardware accelerated video analysis, which provides new opportunities for participants in the IP industry. In this regard, the most important task for security and video surveillance service providers is to follow the speed of change and focus on efforts to achieve highly effective results.
Keywords: Internet of Things, neural networks, video surveillance, machine learning, video analytics.
Cloud and local information services have become fundamental components of modern life, and a new class of services, the Internet of Things (IoT), has emerged, increasing our dependence on network technologies in various areas of human life. IoT services enable communication between everyday devices, such as home appliances, consumer devices, industrial controls, sensors, and virtually anything that transmits information.
Before we begin to study the interaction of the Internet of Things and machine learning environment, we will consider these areas sequentially. The Internet of Things (IoT) is a network of networks where people communicate with their devices, and these devices interact and communicate with each other, respond to environmental changes, and even make decisions without human intervention [5]. IoT devices function independently; however, people can configure them and provide access to data. Internet of Things (IoT) systems operate in real time and consist of a network of smart devices and a cloud platform to which they are connected using WiFi, Bluetooth or other forms of communication. First, the devices collect information, for example, a burglar is found in the house. Then the software processes this information, notifies the user about
it, or performs further actions itself - calls the police. The IoT platform is secure. It has tools similar to those used in Internet banking, namely secure SSL and HTTPS encryption protocols, a network antivirus, and a cyber-threat protection centre. Thus, the platform anticipates equipment wear and possible failures before critical situations occur.
Delving into the future of the Internet of Things (IoT), you can predict that cybercriminals will continue to attack devices, because the IoT system is one of the most reliable and fast ways to spread malware [6]. Users, companies, entire cities and countries are increasingly using smart technologies to save time and money. For example, already now traffic lights with built-in video sensors regulate traffic depending on traffic or, for example, that refrigerators will begin to warn people about the imminent deterioration of certain products. Today, the main problem of IoT implementation is the lack of uniform standards. That is why existing solutions are difficult to integrate with each other, and new ones appear much slower than they could. Also, one of the most important features of the Internet of Things should be autonomy, so that devices can receive the energy of the environment, without human intervention.
