Байєсівська мережа в ІоТ на прикладі побудови кмітливої кур’єрської служби
Loading...
Date
2018
Authors
Кліщенко, Ладислава
Кудрявцева, Юлія
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Розглянуто задачу побудови байєсівської мережі – орієнтованого ациклічного графа, з прорахуванням умовної та безумовної ймовірності змінних на прикладі оцінювання рейтингу водія за
допомогою OBD-II давача.
Bayesian network is an acyclic oriented graph in which each vertex (node of a network) represents an n-valued variable, where arcs mean the existence of direct causal relationships between variables, and the strength of these variables is quantified in the form of conditional probabilities. Bayesian networks are useful for processing data with uncertainties. In this article we define the necessary variables and build relations between them. Also we show how to count the probability of depending variables according to side factors. The information to consider can be defined by experts or based on the previous experience. Our structure of the Bayesian network represents the structure of competences of the driver. Nowadays, many companies are based on reputation systems. In online marketplaces, “reputations” are typically calculated from the numerical feedback scores left by past trading partners [5]. However, according to human psychology, it becomes uncomfortable for people to put a low grade for any kind of facilities. Therefore, it is important to get an objective system for rating counting. We have designed a Bayesian network that provides us with a method of getting a score for driving experience. Getting raw data from OBD-II and analyzing it can help us to estimate the driver. Such a system can be used in any industry where we take care about the quality of driving: taxis, personal drivers, truckers, etc. Creating a network should begin with the definition of variables. We define the classes of these variables. Depending on their role, they could be Target variables, Observations, Factors, and Auxiliary variables. Then we can build relations between them, using directed ribs. The meaning of the rib lies in the fact that the variable in the initial vertex directly affects the change in the target. Then, while counting the probabilities, we must include all possible factors. Finally, the main purpose of the network is to define the right components and the relations between them and to count conditional and unconditional probability for each variable. In our case, we have described the general scheme of analyzing data and working with the algorithm of building the Bayesian network. In conclusion, the purpose is to get the objective driver rating and implement it in other structures.
Bayesian network is an acyclic oriented graph in which each vertex (node of a network) represents an n-valued variable, where arcs mean the existence of direct causal relationships between variables, and the strength of these variables is quantified in the form of conditional probabilities. Bayesian networks are useful for processing data with uncertainties. In this article we define the necessary variables and build relations between them. Also we show how to count the probability of depending variables according to side factors. The information to consider can be defined by experts or based on the previous experience. Our structure of the Bayesian network represents the structure of competences of the driver. Nowadays, many companies are based on reputation systems. In online marketplaces, “reputations” are typically calculated from the numerical feedback scores left by past trading partners [5]. However, according to human psychology, it becomes uncomfortable for people to put a low grade for any kind of facilities. Therefore, it is important to get an objective system for rating counting. We have designed a Bayesian network that provides us with a method of getting a score for driving experience. Getting raw data from OBD-II and analyzing it can help us to estimate the driver. Such a system can be used in any industry where we take care about the quality of driving: taxis, personal drivers, truckers, etc. Creating a network should begin with the definition of variables. We define the classes of these variables. Depending on their role, they could be Target variables, Observations, Factors, and Auxiliary variables. Then we can build relations between them, using directed ribs. The meaning of the rib lies in the fact that the variable in the initial vertex directly affects the change in the target. Then, while counting the probabilities, we must include all possible factors. Finally, the main purpose of the network is to define the right components and the relations between them and to count conditional and unconditional probability for each variable. In our case, we have described the general scheme of analyzing data and working with the algorithm of building the Bayesian network. In conclusion, the purpose is to get the objective driver rating and implement it in other structures.
Description
Keywords
байєсівська мережа, орієнтований ациклічний граф, умовна ймовірність, безумовна ймовірність, навчаємий глибинний аналіз даних, оцінювання водіїв, стаття, Bayesian network, oriented acyclic graph, conditional probability, unconditional probability, IoT, Educational Data Mining, drivers evaluation
Citation
Кліщенко Л. Д. Байєсівська мережа в ІоТ на прикладі побудови кмітливої кур'єрської служби / Кліщенко Л. Д., Кудрявцева Ю. В. // Наукові записки НаУКМА. Комп'ютерні науки. - 2018. - Т. 1. - С. 54-57.