A Novel Approach to Short-term Stock Price Movement Prediction using Transfer Learning

웬 티 투 Department of Electrical and Computer Engineering, 2020 국내석사

Stock price prediction has always been an important application in time series predictions. Recently, deep neural networks have been employed extensively for financial time series tasks. The network typically requires a large amount of training samples to achieve high accuracy. However, in the stock market, the number of data points collected on a daily basis is limited in one year, which leads to insufficient training samples and accordingly results in the overfitting problem. Moreover, predicting stock price movement is affected by various factors in the stock market. Therefore, choosing appropriate input features for prediction models should be taken into account. To address these problems, this thesis proposes a novel framework named deep transfer with related stock information (DTRSI), which takes advantage of a deep neural network and transfer learning. First, a base model using long short-term memory (LSTM) cells is pre-trained based on a large amount of data, which are obtained from a number of different stocks, to optimize initial training parameters. Second, the base model is fine-tuned by using a small amount data from a target stock and different types of input features (constructed based on the relationship between stocks) in order to enhance performance. Experiments are conducted with data from top-five companies in the Korean market and the U.S market from 2012 to 2018 in terms of the highest market capitalization. Experimental results demonstrate the effectiveness of transfer learning and using stock relationship information in helping to improve model performance, and the proposed approach shows remarkable performance (compared to other baselines) in terms of prediction accuracy.

Spectrum-Aware Routing in Cognitive Ad Hoc Sensor Networks

가푸르 후마 Department of Electrical and Computer Engineering 2018 국내박사

The earth, a watery place because of a higher percentage of water than the land, has several communication technologies that help to improve our living standards. Observing our planet in this regard, there exist different living organisms on the land surface, the sea surface and in the ocean which communicate daily with their species using different spectrum. Along with these living things, there are several non-living things with some similar characteristics of living things on the earth e.g., vehicles that need to be fed with petrol, sensors that have ability to adapt according to the environment and need energy to continue existing, machines or robots that can move, and many more, which utilize the same spectrum for communications. Nowadays, we are more than 60\% dependent on these non-living things to make rapid advancements in inter-networking. Inter-networking is a connecting phenomenon that requires a routing protocol to transfer the data packets between different networks by using gateways. Routing is a process that helps sensor nodes to establish a stable link to forward a message to its destination. Hence, to improve communications among different kinds of communicating devices on the land, the sea surface, and in the ocean, three types of sensor networks: terrestrial, maritime, and underwater are designed to deal with various applications, respectively. As we are exposed to a plethora of mobile applications over the past few years, our living standards are becoming increasingly the part of smart networking. Among various kinds of other systems that are essential to improve our living standards all over the world, the intelligent transportation system is the one that overcomes serious issues due to road accidents. Millions of deaths are caused by road accidents. Therefore, in this thesis, we consider vehicular ad hoc networks as the terrestrial networks. Vehicular ad hoc network is a promising mean for safe driving by enabling cooperation among vehicles. And when it comes to safe and stable communications at the sea surface, maritime ad hoc networks are the ones that play an essential role in providing a variety of safety to users aboard. Similarly, communications in the ocean have also been attracting significant interests to deal with various applications for underwater networks. Hence, in this thesis, we consider three different types of communications systems: vehicular ad hoc networks, maritime ad hoc networks, and underwater acoustic networks to deal with the developing requirements of their applications by ensuring safe and stable communications; and intend to overcome the existing issues in each of them. Both vehicular and maritime ad hoc networks use electromagnetic radio waves as a medium of communications, whereas underwater acoustic networks use acoustic waves. Ubiquitous wireless communications is an essential goal for numerous applications ranging from traffic safety to entertainment-related information for various users either on the land, the sea surface or in the ocean. The dedicated licensed spectrum for each of these communications systems has been found insufficient to fulfill the increasing needs of vehicular, maritime, and underwater applications. To alleviate the spectrum scarcity in these networks, cognitive technology is a viable solution as it can utilize spectrum in an environment-friendly manner (i.e., avoiding harmful interference with licensed users). To this end, stable links are essential for communications with different users in order to meet the growing demands of vehicular, maritime, and underwater applications. A link is formed only when two communicating nodes have consensus about a common idle channel. Therefore, novel cognitive routing protocols are required for each of these networks to ensure cooperation among the respective users; thereby retaining stable links for vehicular, maritime, and underwater communications. Various routing techniques have been proposed for vehicular, maritime, and underwater networks, but the number of routing protocols that consider cognitive capability with a routing technique is very limited for vehicular networks. Nevertheless, safe and stable communications issues for cognitive vehicular networks are still under investigation in order to reach a robust and distinguished solution. Similarly, for maritime and underwater networks, combining cognitive principles with routing schemes have not yet been considered. Therefore, in this thesis, we first propose cognitive routing protocols that ensure stable routes between sources and destinations in order to overcome the problems of spectrum scarcity and high latency in vehicular, maritime, and underwater networks, respectively. Our goal is to maintain network stability by considering spectrum sensing and routing simultaneously for vehicular, maritime, and underwater communications. We prove better network performance in each of these cognitive routing protocols in terms of end-to-end delay, delivery ratio, and routing overhead. From these results, we observe that the performance of these networks can be further improved by considering a logically centralized controller that has a global view of the network states and is responsible for selecting the stable paths. This is only possible with the physical separation of network control plane and the forwarding plane. Therefore, we then apply a new concept of software-defined networking (SDN) in these cognitive vehicular, maritime, and underwater networks to further overcome the shortcomings with the existing architectures in these domains. We find that the SDN-based cognitive routing protocols for each of these vehicular, maritime, and underwater networks improve network performance in comparison with non-SDN-based cognitive routing protocols. All nodes in non-SDN based networks perform all functions (routing, forwarding, and network management) individually resulting in an inefficient utilization of resources, high latency, and large amounts of overhead. Due to SDN approach, these nodes do not further need to configure individually. Any change in the network can be now done centrally by the logically centralized controller. We further comprehend from these results that the improvement is only for networks with specific applications. In order to support multiple applications simultaneously under the same infrastructure and enable users to satisfy each application service with improved network flexibility, we finally introduce two integrated architectures. The first one supports different vehicular applications with the integration of software-defined networking, network function virtualization, and fog computing. However, the second one is an integrated coastal city that instate cognitive vehicular-to-ship communications in hybrid environments. Consequently, we end up this thesis opening a new door for routing in integrated cognitive vehicular and maritime networks in order to run multiple applications simultaneously under the same infrastructure.

MACHINE LEARNING BASED TIMELY DETECTION AND DIAGNOSIS OF ABNORMAL BEHAVIORS IN THE INTERNET-OF-THINGS

잔 사나 울라 Department of Electrical Engineering, University o 2020 국내박사

The Internet-of-Things (IoTs) is believed to address social issues and challenges of the modern world to a large extent. For instance, the issues faced by big cities, such as increasing rate of unemployment, economic downfall, the air pollution due to emissions from vehicles or industries, useless energy consumption in homes, or the diculties experienced by handicapped and elderly people, and so on. The IoTs provides opportunities to develop applications and platforms that can be used to regulate and reduce the factors that cause such problems. In addition, the IoTs posses a great potential of providing opportunities to developing new applications that improve the quality of life of citizens. The concept of IoTs is to interconnect the electronic devices (referred as things in the terminology of IoTs) through a cloud network to perform actions without human interaction. These devices include sensors at one end of the network to sense and collect information about the environment. A communication medium with specic protocols is used to transfer this data from sensing nodes to the central units. These units process thecollected data to extract useful information and execute a command or action accordingly. This idea of interconnected things with ability to act independently is approaching toward reality with the increase in number of nodes connected to the internet. The ubiquitous sensing nodes, or simply sensors, deployed in IoTs network are mobile or installed at xed positions spread over a large area. These nodes rely on wireless communication technologies to transfer the collected data to the central unit through cloud platform. Moreover, these nodes are constrained in terms of computational resources, memory and battery to store energy making it impossible to implement the existing computationally expensive cybersecurity techniques used in traditional Wireless Sensor Networks (WSNs). Therefore, the nodes of IoTs are more prone to cyber attacks as compared to the constraintfree immobile nodes of WSNs. This situation is posing a serious challenge of security and data privacy while implementing the IoTs network. Furthermore, the performance of IoTs is highly dependent on the sensed data collected by sensors. A failure originated in these nodes (also known as fault) may corrupt the collected data leading to serious consequences, such as huge economic losses, long delays in the network function, or human safety in some cases. Therefore, it is necessary to develop algorithms for detection or prediction of the fault occurrence in IoTs nodes. In short, the IoTs network should be capable of detecting faults and intrusions in order to implement a reliable network free of disturbances. However, the resource limitation problem of the IoTs nodes and the fact that these nodes are deployed in the remote areas where the environmental conditions may vary rapidly, put limitations on the development of faults and intrusion detection mechanisms. Firstly, the algorithms designed for such sensors should perform well with the given amount of resources in the node. Secondly, these algorithms should be capable of distinguishing the faults or attacks from the behaviors occured due to variations in the environment. This dissertation is related to developing the fault and intrusion detection algorithms for IoTs nodes while consideringthese challenges. Firstly, a fault detection and diagnosis framework is proposed to identify the presence and type of fault originated in sensors. A well-known data-driven supervised machine learning-based classication algorithm, called Support Vector Machine (SVM), is utilized to design this framework. The classier is given input in terms of statistical time-domain features extracted from the raw signal of sensor. The dataset used to analyze the performance of the classier is acquired as follows: normal data signals are obtained from a temperature-to-voltage converter TC1047/1047A by using an arduino Uno microcontroller, and a personal computer with MATLAB. Then, ve types of faults, including drift, gain, precision degradation, spike and stuck fault, are injected in the normal data signals. The performance of the proposed algorithm is presented under dierent scenarios. Furthermore, a comparison with neural network-based algorithm show that the proposed algorithm perform better. Secondly, a novel feature selection scheme is proposed to select most discriminating subset of features among the whole set of features. The features having high mutual information with target class and minimum mutual information with non-target class are selected. This method is veried using the similar dataset. Thirdly, a distributed fault detection and diagnosis approach is proposed using stacked auto-encoder, SVM, and fuzzy deep neural network. The fault detection can be implemented in the resource-constrained sensing nodes to avoid the delays occured due to data transmission between the sensor and the central unit. Nevertheless, the fault diagnosis can be carried out in the central unit to obtain more information about the detected fault. Again, this technique is veried using the similar data with the same types of faults. Moreover, a lightweight intrusion deteciton system is proposed using SVM-based classier with three non-complex features, namely mean, maximum, and median. The denialof- service (DoS) attacks are targetted which have a concomitant increasing or decreasingeect on the trac intensity. The anomaly-based intrusion detection system continuously monitors the trac intensity attribute of the node to detect intrusion. The performance of the proposed system is anayzed using a dataset obtained using simulations with dierent types of DoS attacks, including packets ooding, vulenerability, blackhole, jamming, selective forwarding, sybil, sinkhole, clone, wormhole, and hello ood attacks. Finally, an SVM-based spectrum sensing technique is proposed to detect the presence of primary users in cognitive radio networks. The secondary users start transmission if the channel is sensed idle or free from primary user with a transmission power calculated based on the quantized sensing result and the residual energy in node battery. The objective is to improve the throughput performance of the system. The simulation results shows that the proposed algorithms successfully achieve the objective of improved throughput as compared to the conventional energy detection technique.

Highly Sensitive Detection of Influenza A (H1N1) Virus with Functionalized-gate FETs

최원영 Department of Electrical Engineering Pohang Univer 2018 국내석사

We fabricated functionalized-gate field-effect transistors (FGFETs) and demonstrated highly sensitive detection of influenza A (H1N1) virus. Monoclonal antibodies (mAbs) of H1N1 virus were immobilized on the external-gate. The fabricated FGFETs showed excellent intrinsic electrical characteristics. We observed the shift of transfer curve after 10 g/ml of the H1N1 virus is added and extract the sensitivity from drain current change. FGFETs can avoid the drawbacks of the conventional BioFETs that cost inefficiency due to the disposability of the sensor after usage. Therefore, FGFETs are very promising to be used as multi-biosensors.

A Location-based Delay-constrained Task Assignment Framework in Mobile Crowd-sensing

악터 샤티 Department of Electrical and Computer Engineering, 2020 국내석사

Mobile crowd-sensing (MCS) has recently become a promising approach for massive data collection, which empowers common people to perform sensing tasks with their smart devices. In MCS, locations of tasks and workers are diverse, and workers need to visit different task venues to perform the tasks. The diversity of task and worker locations, tasks’ location accessibility, and required sensor type make the task assignment problem highly challenging. In time-sensitive MCS applications, this task assignment problem becomes even more intractable because of the deadline and a lot of possible movement trajectories of the workers. In this paper, we introduce two types of workers and formulate the task assignment problem, which comprises an embedded structure. Furthermore, a decomposition technique is applied to decompose the original problem into the main problem (the assignment problem) and a set of sub-problems (traveling salesman problems). The assignment problem determines task-worker assignments, and the sub-problems determine trajectories of the workers. This decomposition allows using a simpler solution strategy. Then, a memetic genetic algorithm is proposed to address the assignment problem, while each sub-problem is solved using an asymmetric traveling salesman problem heuristic. Results from simulations verify that the proposed algorithm outperforms the baseline methods under various experimental settings.

DEVELOPMENT OF ROBOT MANIPULATOR CALIBRATION TECHNIQUES USING MODEL BASED IDENTIFICATION AND UNMODELED COMPENSATION METHODS

레 푸 응우엔 Department of Electrical Engineering University o 2021 국내박사

In recent years, interest in the application of robot manipulator to automated manufacturing soared. The advent of highly capable computer-controlled manipulators indicate d that tru-ly flexible automation was feasible, and many manufacturers rushed to take advantage of this technology. Robots are used in a wide range of tasks in industrial applications such as material handling, milling, painting welding, and roughing. Although the modeled-based robotic calibration methods have been widely researched for decades. It is difficult to cre-ate models that consider all the causes engendering the end effecter error. Therefore, to archive further accuracy, a good deal of attention has been paid to the area of un-modeled calibration for the sources of errors that could not be taken into account by model-based calibration. In this study, new robotic calibration methods are introduced. By combining the joint deflection model with the conventional kinematic model of a manipulator, the geometric errors and joint deflection errors can be considered together to increase its positional accu-racy. A new method includes the kinematic calibration and non-geometric compensation with a RBF compensator that compensates for compliance errors based on the effective torques. To improve the effectiveness of the calibration process, a neural network is de-signed to additionally compensate the unmodeled errors, specially, non-geometric errors. Then, the weights and biases of the neural network is determined by conventional back-propagation method. For increasing the ability of the neural network, heuristic optimization methods such as teaching learning optimization and invading weed optimization methods are hired for better convergence capability than the back propagation neural network in this calibration process. This work also presents a new method includes the kinematic calibration and teaching learning-based optimization for directly determining joint compliance parameters. The ad-vantages of the suggested method are easy for implementing, removing the need for torque sensors, high ability to enhance the precision of the manipulator. In order to demonstrate the effectiveness of the proposed method, experimental studies are carried out on manipulators. The enhanced position accuracy of the manipulator after the calibration confirms the feasibility and more positional accuracy over the other calibra-tion methods.

(A)New Auto Exposure and Auto White-Balance Algorithm for High Dynamic Range Conditions using CMOS Technology

Vuong, Quoc Kien Department of electrical engineering, Graduate sch 2009 국내석사