Providing a framework to improve the prediction of the traffic situation

Master's Thesis in Computer Engineering (Artificial Intelligence)

Today, the success of intelligent transportation systems depends not only on the information of the current traffic situation, but also to a large extent on the knowledge of the traffic situation in the next few minutes. Therefore, a lot of research has been done in the field of short-term prediction of the traffic situation. However, the emphasis of the majority of them is only on the application of different algorithms in order to learn the traffic data and present the model, based on the data collected from the current and previous situation. However, in order to achieve an efficient algorithm, it is necessary to consider the fluctuating and time-dependent nature of the data in the learning process of the model. In this regard, this thesis, by studying the distribution of traffic flows, tries to separate the behaviors related to peak and non-peak traffic periods, as well as using the concepts and knowledge obtained to teach different models corresponding to different traffic behaviors. It is worth mentioning that even if the time associated with the data is not explicitly available, the proposed method detects the trend of traffic flows by examining the data distribution. In this way, Random Forest as a predictive model is aware of the training data context and accordingly, the probability of getting stuck in local optimality is reduced. In order to evaluate the presented method, experiments were conducted on the data of the traffic section of the 2010 International Data Mining Competition. The results confirm the efficiency and scalability of the proposed method compared to other results obtained by the top teams of the competition. Chapter 1 Introduction Introduction noise, fuel consumption, waste of time and energy and their imposed costs, providing a suitable solution for smooth traffic flow is of particular importance. On the other hand, due to the limitations of urban development facilities in front of the mass demand of vehicles, it is necessary to consider practical and possible measures to solve this problem. Since information technology [1] has played an effective role in various industrial fields, the introduction of this technology in the field of transportation systems was also considered as a suitable solution and led to the emergence of intelligent transportation systems [2]. In fact, information technology technology allows the elements of the transportation system to become an intelligent system by using sensors [3] and microchips and their communication through wireless technology [4]. Today, the intelligent transportation system has taken an effective step towards the management of the transportation system and intelligent use of the existing infrastructure by forming a system consisting of data receiving sensors, information processing systems, and information presentation systems to users [1]. For example, this system has gained control of this field by using different technologies such as vehicle guidance and traffic light control system, traffic notice boards, speedometer camera and automatic vehicle number identification system to advanced and more complex systems that simultaneously integrate different information such as weather conditions, traffic conditions, road conditions from different sources. Among the important achievements of using the smart transportation system, we can mention the reduction of traffic, the reduction of accidents and accidents, the possibility of choosing optimal routes according to the condition of the routes, the management of public transportation and emergency vehicles, as well as the possibility of electronic collection of things such as tolls, parking fees, and ticket purchases, which lead to saving fuel and energy and reducing imposed costs. Generally, intelligent transportation systems are examined under the heading of five main groups, each of which includes different areas of this system;

A) Advanced Travel Information Systems[5] (ATIS) whose task is to provide information on the current traffic and weather conditions of the roads, accidents and road repairs, as well as inform passengers and users in order to optimally use the existing routes and establish traffic balance.

b) Advanced traffic management systems [6] (ATMS) which examines and integrates traffic information collected from various sources and through traffic control tools such as traffic signals, ramp control [7] at the entrance of highways in order to maintain density and variable information boards on the roads.

B) Advanced traffic management systems[6] (ATMS) which examine and integrate traffic information collected from various sources and control traffic flow through traffic control tools such as traffic signals, ramp control[7] of highway entrances in order to maintain density and variable information boards on the roads.

C) Electronic payment systems[8] (EPS) which includes the system Electronic toll collection [9] (ETC) is a toll payment system for the use of special lanes for high-passenger vehicles [10] by single-passenger vehicles, as well as route pricing [11] and high-traffic lines.

d) Advanced and intelligent public transportation systems [12] (APTS) are things to facilitate the provision of public transportation services such as determining the automatic location [13] of the vehicle and informing It also includes passengers, reservation and fare determination services.

e) Advanced Vehicle Control Systems (AVCS) [14] which include Intelligent Speed ??Adaptation System [15] (ISA), warning and accident prevention systems. 

In the field of AITS and ATMS, short-term traffic forecasting is one of the important elements of the success of intelligent transportation systems, because in the direction of traffic control, not only the current traffic situation but also the future traffic situation is important. Therefore, traffic forecasting algorithms have received special attention among researchers in this field. 1-2 Challenges of the problem As stated earlier, traffic control centers make the necessary decisions for traffic management and control based on the collection of traffic statistics and information, their processing and integration. In order to improve traffic control, ATIS and ATMS as the main components of the intelligent transportation system, in addition to the current traffic situation, also need the future traffic situation. Therefore, predicting the future state of traffic is one of the important topics for these centers, so that by using it, the necessary strategies can be used to prevent congestion and warn drivers to choose the optimal route. So far, many researches have been conducted regarding the prediction of the future traffic situation, which actually predict the future traffic using the recorded data of the current traffic situation. Usually, the data collected in the field of traffic are provided to us in the form of time series [16], which actually include different records that are obtained in equal time intervals and during consecutive measurements. Using current and past data, their values ??are predicted in the future [2]. So far, different techniques have been used in the field of traffic forecasting, among them Kalman filtering methods [17] [4,3], non-parametric statistical methods [5,6] [18], sequential learning methods [7] [19], neural network models [20] [8-11] and time series analysis [13-17]. One of the most important challenges of applying these algorithms is the high volume of traffic data, which has led to the recent trend of research towards the use of data mining algorithms[21]. Among them, methods based on decision trees [22] have been widely used in the field of traffic [18,19]. Also, cumulative learning methods [23] such as bagging and boosting were given special attention due to their high efficiency. Their main idea is to build a set of models and combine their results with the aim of improving the accuracy [24] of learning [47]. In figure-11, we see the general architecture of cumulative learning algorithms taken from the book [20]. ABSTRACT A Framework For Improving Traffic Conditions Prediction Nowadays short-term traffic prediction is of great interest in Intelligent Transportation Systems (ITS). To come up with an effective prediction model, it is essential to consider the time-dependent volatility nature of traffic data. Inspired by this understanding, this paper explores the underlying trend of traffic flow to differentiate between peak and non-peak traffic periods, and finally makes use of this notion to train separate prediction model for each period effectively.  It is worth mentioning that even if time associated with the traffic data is not given explicitly, the proposed approach will strive to identify different trends by exploring distribution of data