Українська   Русский
DonNTU   Masters' portal

Abstract

The contents

1. Relevance of the topic

Today humanity is adequately examined the behavior of the individual and crowd behavior, revealed a number of factors that determine their type of behavior and the causal relationships that lead to a variety of events. However, to date, practical developments in this area is not a lot.

Many companies of the country, the field of science are interested in getting answers about the reactions and actions of different people, in response to all kinds of events: the release of new products, conducting meetings, and other emergency. With the successful results of constructing models to predict the crime situation will be possible to predict a wide range of human activities, due to the possibility of introducing new features into the system and edit the expert assessments in accordance with the task.

Even without modification, the subsystem is able to answer questions about the importance of individual events on different people, from the point of view of the possibility of the commission of their crimes, is extremely important. Also, a significant aspect of future subsystem is the ability to specify the most unreliable elements in the city, starting from the characteristics of the particular individual.

2. The purpose and objectives of the research

Purpose of the diploma work – to study the characteristics and specificity of the formation of opinion in humans and human masses, the ability to realize the simulation of events and human behavior, taking as an example the simulation of committing human crimes.

To achieve the goal of the diploma work is necessary to solve the following problems:

The subject of this research is to model the causal link, which leads man to commit crimes of various nature, based on the current status of the individual, as well as its near and distant past.

The object of the study are psycho-physical characteristics of a person and based on their individual activities, methods and tools for building software systems for simulating the simulation of human behavior.

3. Planned practical results

As practical results it is planned to develop a cross-platform, customizable environment for modeling human criminal activities, with the following properties:

  1. Implementation taking into account the opportunities for multiprocessor systems;
  2. The possibility to monitor in pseudo-real (much reduced) time of the simulation results;
  3. The ability to predict the commission of a crime, under certain initial data;
  4. The ability to generate a report on crime with a variety of configurable options;
  5. The ability to predict an individual's response or the crowd at certain events (meetings, presentations, accident, etc.). First of all, in terms of crime.

4. Alleged scientific novelty

To solve the problems of modeling or predicting the behavior used many methods: polls, collecting information on the preferences of the person at his conversion to the search systems in the Internet, statistical analysis of past reactions and other human masses and etc. But all these methods are often insufficient to quality prediction of the individual behavior.

At this moment, no system, modulating the criminal activities of a person based on his past, his social relations and current events. Developed subsystem involves the analysis of human behavior and modeling of committing an action based on many factors.

5. Overview of international sources

The study of cause-effect relationships in the human activities carried out for many years. As of today identified the basic mechanisms that influence behavior of individual, down to specific events and reactions to them are identified. However, software development in this direction are represented in a small amount. Even fewer software products on this topic, created in the field of criminology.

Since the research paper focuses on the modeling of illegal human activities, it should be noted that was created a whole section of scientific knowledge in this area: Criminological forecasting.

Criminological forecasting – a prediction of the future state of crime and related phenomena and factors (criminological environment), as well as identifying the main trends of their development. Criminological prediction is based on available in criminology scientific concepts of crime, using statistical and experimental methods, mathematical modeling. In order for the result can be used in practice, it must have scientific character, possess the property verifiability. [1]

Software created for criminological forecasting, have one common drawback: they are aimed at the definition of a statistical relationship between the state of crime in the preceding and subsequent time intervals. That is, in these systems are calculated quantitative indices of crime for a specific region, based on the same figures for the previous time intervals.

5.1 Overview of international sources

The software DI Guy – module of artificial intelligence, modeling crowd behavior with considering of each individual [2].

The program was developed and presented by VT MAK. The main core of the system, for the simulation of human behavior, is presented in the basic package DI Guy. Additional elements include DI Scenario, DI SDK, DI Motion editor. With their help it is possible to develop new scenarios of behavior and situational environment, to achieve a more realistic behavior of the individual (such as a kicking motion or walking in circles while waiting); add mapping facial animation in a 3D format.

Embedded in the program artificial intelligence and graphics capabilities make it possible to create realistic street fights, rallies, military operations, interrogation of suspects and many other situations. In connection with the well-developed capabilities for modeling the behavior of people during the war, a great interest in this project show the US military. Moreover, it is important to note: for simulation of collision in a peaceful area of the city with the enemy soldiers, you can simulate the behavior not only of the soldiers, but also the behavior of civilians (each individually), as well as different animals (chickens, cows, horses). In a qualitative simulation offers the possibility of training soldiers to identify potential enemies by gestures and facial expressions, even taking into account the behavior of other people's culture. [3]

The result of the simulation by using DI Guy

Figure 1 – The result of the simulation by using DI Guy
(Figure created from VT MAK website screenshots, page DI-Guy SDK [2])
(animation: 6 frames, 7 cycles of repeating, 179 kilobytes)

MOVES Institute at the US Naval Higher School in Monterey, California, is engaged in modeling human behavior. Since this is a military school, the main direction of modeling - military operations and people's reaction to them. However, further development is aimed also at the social and cultural modeling. Of the specific of the institution would like to mention the FOCUS project, led by Steve Hall and Jeff Applegeta and Cognitive and Perceptron Modeling . The first draft of the model allows you to build a social network population that adequately responds to media work and helps to determine the type of broadcasting, in which the population will be configured as loyal to the army. This development will help to create preventive program for reducing the number of terrorist attacks in the occupied city , and to reduce the crime rate. In addition, it is planned to create the functionality to predict the most unreliable (hazardous) locations. The second project is created to improve understanding of how the information that is available to an individual, can be integrated into the likeness of a human mind to predict the consequences of various courses of action. That is, the program will be able to give an answer on how a person perceives a certain activity and the possible actions of a person, in response to any action. [4]

Another interesting project is the project of Marco Strano and Robert Bruzon. They have assembled a team of psychologists, medical experts and investigators. The purpose of the commands has been the development of the statistical model that links criminal profile with his / her actions and decisions, as well as about the consequences of the crime scene.

Marco Strano and Robert Bruzon defined variables that play a significant role in the characterization of the behavior of the offender during a violent crime. One hundred and fifty of these variables were selected by analyzing the literature and by interviewing a sample of several hundred investigators and medical experts from different countries, who are usually involved in the murder investigation. These variables are used to construct the so-called Bayes Networks. Bayesian networks are graphical models that extract the knowledge of the system of empirical data and indicate a causal relationship between all relevant variables. Using conditional probabilities, they can determine the degree to which variables can influence each other, even if the underlying mechanisms are unknown activity. This work aimed to determine the profile of the offender and the simulation of its possible behavior. [5]

5.2 Overview of national sources

The project model of human behavior - is an informal community of researchers involved in some form of computer modeling of human behavior [6].

The project is a modular program, with the possibility of inclusions or disable major components. The components are emotional module, experience module of the environment, energy balance module, addiction and automatisms and others. The central element of the model - a man. The second element of the model is the importance of human activities, meaning its action. Currently implemented are not a lot of action, but, right now, there is a certain set of alternatives for each action. So I get the problem of decision-making modular man. To solve this problem factors (time, energy), goals, emotions, interests, and the results of actions have been introduced and others.

It is worth noting: In addition to the simulation of the individual, the model suggests modeling groups of people interacting with each other. It was introduced for a more accurate description of the man, because man is social. Also there are companys in the present model, in which people can work.

By today released a demo version of the model of human behavior, which can be downloaded here.

Tool AnyLogic simulation was developed by a Russian company The AnyLogic Company. This software environment is created using various methods simulation. Combining system dynamics, agent-based and discrete event simulation can be used in a software package AnyLogic different fields. [7]

The project includes various libraries written in Java. These libraries implement certain system features, such as the simulation of the physical environment, imitation yard operations, interactive animation 2D and 3D models, simulation of car traffic and much more.

The use of agent-based modeling allows AnyLogic to simulate the behavior of human masses. At the moment already have some experience in this field, for example: Simulation of crowd behavior [8].

Theoretical studies in the CIS, in the direction of the theme of the diploma work is quite extensive. For example, in the work of Loginov Eugene Aleksandrovich [9] reveals the potential of criminological prediction. In this paper we present the basic preconditions for the consideration of criminological prediction as the basis for identifying the most unreliable citizens and potential offenders. Also, Eugene is argue for favor of the study of individual criminal behavior on the basis of criminological prediction.

One of the most promising ideas on the theme of master's work, put forward in the article, is forecasting model based on assumptions Minin AY This model is constructed using the correlation of a person with a specific prototype. Allocation takes place on a predefined set of features.

Another article that is worth mentioning is the work of A. Miller [10]. According to the article, it is considered to be objects of criminological prediction: ... Personality of the offender - taken in the development, in conjunction with external factors, a set of social features and characteristics, and anti-social orientation of the individual plants, the effects of which society and caused them to commit a wrongful socially dangerous act. (Excerpt from an article by A. Miller [10]). Criminal behavior in the article is divided into three parts: before the criminal behavior, criminal behavior and post criminal behavior. The very same prediction, is divided into a retrospective and prospective analysis of the simulation. The article clearly points to the possibility of applying criminological prediction as a tool for modeling the offender and his behavior.

5.3 Review of local sources

In the Donetsk National Technical University, in the direction of the master's work, there are several master's works.

The problem criminology studied Zenkevich Kirill [11]. His work has been aimed at drawing up an expert system to determine the best methods of individual work with special contingent. This paper describes many aspects of the psychology of criminals.

Modeling of human behavior with the help of agent-oriented programming was engaged Lukin Julia [12]. She is focuses on the use of MadKit program, in which it is possible to build multi-agent system and which it was planned produce a simulation of the behavior of human masses.

Description motivational device of the person engaged in Poveshchenko Tatiana [13]. Despite the narrow specialization of her work (labor behavior), Tatiana in some detail and correctly described the general concept of human motivation is the mechanism of the human psyche.

Socially-oriented multi-agent model developed Stropalov Andrew [14]. The main part of a master's site, he has devoted a general description of the multi-agent programming, but his work involves the development of socially-oriented system, simulating the behavior of large a human masses.

The consideration of methods to the development of agent-based systems involved Pocpelov Sergey [15]. His work was based on the problem of the control an autonomous agent in order to further the use of artificial intelligence in computer games (for the creation of a separate character).

6. Brief theuretikal foundations

Currently still under analysis methods for constructing models. The emphasis is on the use of evolutionary programming methods or neural networks, or agent-based modeling. The greatest potential for writing subsystem is seen in the use of simulation. Despite the final method for constructing models, theoretical basis for modeling one person have been already developed and will be applied. Theoretical bases at this stage is quite bulky, since trying to more fully describe the behavior of a person. For brevity's sake, explain key concepts:

Events play a major role in human behavior: they can form the majority of indicators of human and some of the events themselves are characteristics. It can be said that the events provoke other events.

There is a certain set of characteristics in humans. Considering them and Maslow's (figure 1) hierarchy of needs [6], formed the current set of needs.

Maslow's hierarchy of needs

Figure 2 – Maslow's hierarchy of needs
(Figure taken with Psihologos site pages Maslow's hierarchy of needs [11])

On the basis of current needs the desires are formed. Simultaneously a person can perform only one action, so the order in which will take into account the needs and desires is crucial. (First calculated the chance of committing actions related to the lower layers of the pyramid, and then, if they have not committed, climb up) [7-8].

Similarly, human actions depend on the subcategory of its characteristics, called "emotions and state". They are formed partly by chance and partly on the basis of the latest developments. Their role are in the application of effects on humans. For example: person has all turns out today – the emotion of happiness, affects a person's mood, it seems that everything will turn out, and he can not commit the planned crime [9].

7. The formalization of the problem

To formalize the problem was introduced dataset [10] works with the subsystem:

Here is an example of formalizing dependencies between the data.

To calculate the chance of events drawn up by the desires of the table of values, which represent the characteristics of the line, and columns – ways to achieve. The cells are entered four values established by experts: the average rate and the norm in the negative and positive, of this figure, the parties (in the table the value displayed on the chance of an increase), as well as the weight indicator.

The weighting factor lies in the range 0 to 2. This range is selected to be able to how to reduce or increase the significance criteria, however, the test unit has thus of great importance.

The data can be completely diverse, so they must be reduced to a common value. The following algorithm has been designed with this purpose:

  1. On the basis of the generated table, the current values are calculated for all cells. This occurs for two formulas for the range [-1, 0] and [0, 1], respectively:

    Xk=-1+(x-a)/(b-a)

    Xk=(x-b)/(c-b),


    where x – current value, a – set minimum value, b – the average rate, c – a predetermined maximum value.
  2. The calculated value Xk is substituted in the function for the relevant range [-1, 0] and [0, 1]:

    Z=-Xk^2 (Z=Xk^3)

    Z=Xk^2 (Z=Xk^3)

    The following graphics of functions, display data conversion (figures 2 and 3). The idea is that a small deviation from the weighted index, significantly less in importance than large deviations. For example, the deviation of the average wage by 10% in the smaller side, can call no reaction at all – man can not be taken into account at the time, as a deviation of 70% can lead to a huge number of reactions, including the crimes. Based on this, we can conclude that the changes are not linear. These amendments are most correctly described by the parabola y = x ^ 2 and y = x ^ 3, the choice of one of them is possible only if the final run of the program and to identify the most appropriate results.

    Charts functions after the data conversion for negative values

    Figure 3 – The graph of the function for negative data reflecting calculated in the first paragraph of the value unchanged (Y = x, green), and the value obtained by converting the data with the function Y = x ^ 2 (blue), and Y = x ^ 3 (red)

    Charts functions after the data conversion for positive values

    Figure 4 – The graph of the function for positive data reflecting calculated in the first paragraph of the value unchanged (Y = x, green), and the value obtained by converting the data with the function Y = x ^ 2 (blue), and Y = x ^ 3 (red) positive values

  3. The resulting value Z is multiplied by the weight factor:

    Zk=Z*w.

  4. Zk values are summarized in columns (for each individual to achieve a method):

    S_i=Σ(Zk)

  5. Total amounts add up and the odds are determined by each method to achieve:

    P=S_i/Σ(S_i)

  6. Generate a random value in the range [0, 1] and P modeling chances for a full group of events, issued a decision on selecting an appropriate method to achieve.
  7. To make the final selection of whether or not the action, another generated random value in the range [0, 1]. And if it is greater than 2/3 (4/5), then it is done, otherwise it is not done.

The chosen method takes into account the maximum number of characteristics. In the future, on the basis of this method, it is planned to develop a complete model of the individual and the creation of models of interaction between groups of individuals to imitate their joint activities.

Conclusions

As a result of research on the topic of the diploma work, the modern concept of human motivation have been studied; are revealed the factors affecting the various groups of events; collected and classified human basic characteristics. The distinctive features of the existing systems of this kind were identified and the uniqueness of the developed subsystems have been defined. Methods and algorithms have been investigated to implement a certain portion of subsystem and a proprietary algorithm for simulation was designed.

References

  1. Долгова А.И. Криминология: Учебник для вузов / Под общ. ред. А.И. Долговой. 3-е изд., перераб. и доп. М.: Норма, 2007. – 415 с.
  2. VT MAK – Project Humans – DI-Guy SDK [Электронный ресурс]. – Режим доступа: http://www.mak.com/products/humans.
  3. Макаров О. Игра в людей: Цифровой аналог человечества [Электронный ресурс]. – Режим доступа: http://www.popmech.ru/technologies/10068-igra-v-lyudey-tsifrovoy-analog-chelovechestva/.
  4. Chris Darken: Human Behavior Simulation [Электронный ресурс]. – Режим доступа: https://www.movesinstitute.org/research/human-behavior-simulation/.
  5. Marco Strano and Roberta Bruzzone, Modeling of Human Behavior in Violent Crimes [Электронный ресурс]. – Режим доступа: http://citeseerx.ist.psu.edu/viewdoc/....
  6. Текущее состояние модели поведения человека [Электронный ресурс]. – Режим доступа: http://www.humanmodel.ru/model/....
  7. Инструмент имитационного моделирования AnyLogic: обзор [Электронный ресурс]. – Режим доступа: http://www.anylogic.ru/overview.
  8. Бекларян А. Л., Акопов А. С. Моделирование поведения толпы на основе интеллектуальной динамики взаимодействующих агентов / «Бизнес-информатика» / № 1 (31) / 2015 / с 69-77
  9. Логинов Е.А. Возможности криминологического прогнозирования индивидуального преступного поведения / «Психопедагогика в правоохранительных органах» / № 4 (55) / 2013 / с 77-82
  10. Миллер А. В. Личность лица, совершившего преступление и его поведение как объекты индивидуального криминологического прогнозирования / «Наука и современность» / № 1-3 / 2010 / с 208-213
  11. Зенкевич К.А. Разработка экспертной системы определения оптимальной методики для индивидуальной работы со спецконтингентом в условиях ДонСИЗО [Электронный ресурс]. – Режим доступа: http://masters.donntu.ru/2009/kita/zenkevich/diss/index.htm.
  12. Лукина Ю.Ю. Агентно-ориентированные программные модели поведения человека в социально-экономической среде [Электронный ресурс]. – Режим доступа: http://masters.donntu.ru/2011/fknt/lukina/diss/index.htm.
  13. Повещенко Т.С. Мотивация трудового поведения работников [Электронный ресурс]. – Режим доступа: http://masters.donntu.ru/2013/iem/poveshchenko/diss/index.htm.
  14. Стропалов А.С. Нейросетевые модели программных агентов в социально-ориентированных мультиагентных системах [Электронный ресурс]. – Режим доступа: http://www.masters.donntu.ru/2012/fknt/stropalov/diss/index.htm.
  15. Пocпeлoв С.М. Моделирование поведения объекта методами нейродинамического программирования [Электронный ресурс]. – Режим доступа: http://masters.donntu.ru/2012/fknt/pospelov/diss/index.htm.
  16. Козлов Н.И. Пирамида потребностей Маслоу [Электронный ресурс]. – Режим доступа: http://www.psychologos.ru/articles/view/piramida_potrebnostey_maslou.
  17. Maslow A.H. Motivation and Personality. — New York: Harpaer & Row, 1954.
  18. Anne-Marie Feyer, Ann M. Williamson. Human factors in assident modelling / Encyclopeadia of occupational health and safety, 4 edition.
  19. Maurice Allais. Le comportement de l'homme rationnel devant le risque: critique des postulats et axiomes de l'ecole americaine // Econometrica, April 1953, v.21, no.2, p.503–549
  20. Леонтьев А.Н. Потребности, мотивы и эмоции / Леонтьев А. Н. – М.: Изд-во Моск. Ун-та, 1971. – 40 с.
  21. Сменцарев Г.В. Методы формализации в решении задач автоматизированного анализа антропосоциокультурных систем / «Прикладная информатика», том 10, №5(59), 2015 г.