Electronics and Computer Engineering
Permanent URI for this collection
Browse
Recent Submissions
Item SECURE DATA CLASSIFICATION AND MOBILITY MODEL FOR CLOUD DATA GOVERNANCE(I.O.E. Pulchowk Campus, 2021-08) PAUDYAL, RAMESHIn the digital era, cloud computing has emerged as a successful computing paradigm which proposes the on-demand delivery of IT resources such as: computing power, storage and database over the internet. The adoption of cloud computing has created operational and security challenges. Cloud data governance security checklist and parameter is an essential element for measuring the computing security in a cloud. It also helps to manage data with proper security measurements and identify the security requirements. Handling all data with the same level of security measurements is not a sustainable security solution. Data centric security solutions offer sensitivity based security requirements during the mobility of data. In this thesis, a fuzzy logic based data classification model has been proposed for security management and mobility that uses cloud data governance security parameters confidentiality, integrity and availability as an input features. In addition, this model automatically identifies the appropriate security algorithm on the basis of sensitivity level, i.e. confidential, sensitive and public classes. This integrated method of data classification and securing significantly improves cloud data mobility.Item Polyword net: A word sense disambiguation specific word net of polysemy words(Institute of Engineering, Electronics and Computer Engineering, 2021) Dhungana, Udaya RajThis dissertation presents a new lexical resource which is named as 'PolyWordNet'. The PolyWordNet mimics the way how the senses of polysemy words and their corresponding related words are organized in a human mind. A related word of a sense of a polysemy word in a given context is a word that can disambiguate the meaning of the sense of the polysemy word in that context. The rationale behind the organization of words in PolyWordNet is that any simple sentence, which contains a polysemy word, also contains at least a related word (s). A sense of a polysemy word and its related word(s) in a sentence, therefore, have a strong semantic relation which can be used to disambiguate the sense of the polysemy word. Utilizing this semantic relation, PolyWordNet organizes the senses of polysemy words based on their corresponding related words. The organization of words in PolyWordNet is completely different as compared to the existing other popular lexical resources such as dictionary and WordNet. The words in a dictionary are organized based on the alphabetical order. Therefore, the words that spell alike come together but the words with similar meaning get scattered in the dictionary. In WordNet, the words with similar meaning are placed together based on the synonymy set. The polysemy words are the big problems in Natural Language Processing tasks since they create the ambiguity. No lexical databases deals with the organization of words based on these polysemy words. Therefore, the PolyWordNet is developed. The words in PolyWordNet are organized in such a way that the senses of polysemy words and their corresponding related words come together and form clusters. The results obtained from 63 runs of experiments performed on 3,541 words and tested by 4,105 different test sentences show the word organization of PolyWordNet is better for word sense disambiguation. These results also indicate that the word organization of PolyWordNet is acceptable and valid with reference to the popular lexical database- WordNet.Item Analysis, modeling and evaluation of service provider legacy network migration to software-defined IPv6 network(Institute of Engineering, Electronics and Computer Engineering, 2021) Dawadi, Babu RamAvailable with full textItem “Design of Stock Trading Agent Using Deep Reinforcement Learning”(IOE Pulchowk Campus, 2022-09) Lal, Janak KumarThis study adopts Double Deep Q learning algorithm to design trading strategies to trade stocks of four commercial banks listed in NEPSE. The reinforcement learning agent takes discrete actions and gets negative or positive reward from the environment. CNN is utilized to form the policy network. A target network is used to mitigate instability due to Deep Q Network. The concept of experience replay is used to randomly sample the batches of experience from the memory and train the network. The performance of Double Deep Q learning agent was compared with various baseline trading strategies in terms of annualised expected trade return. The maximum annualised expected trade return obtained with traditional baseline methods was 103% for testing data of NABIL, while for the same data the reinforcement learning agent using double deep Q learning algorithm obtained annualized expected trade return of 114.44%. The experiments showed that, Double Deep Q learning agent with experience replay had higher annualised expected trade return compared to baseline trading strategies.Item Sentiment Analysis and Topic Modeling on News Headlines(I.O.E. Pulchowk Campus, 2022-09) Yadav, VijayIn today’s world of competitiveness, sentiment analysis has wide range of applications from medical perspective to examine the mental situation of the person to entertainment industry, corporates, politics and so on in order to examine the perspective and views of the people towards their product. News media play vital role in shaping the views of public regarding any product or people. The dataset used for this thesis is headlines dataset of one of the leading new portals of India i.e., Times of India. Both supervised and unsupervised techniques would be used to perform the analysis on the dataset. The thesis has two aspects i.e., first, sentiment analysis for which supervised technique Bi-LSTM will be used and second, topic modeling for which unsupervised techniques LDA (Latent Dirichlet Allocation) and LSA (Latent Semantic Analysis) will be compared, and then the best performing algorithm will be used for topic classification. The topics identified will be used to classify the dataset so that prediction of topic for particular headline can be done.Item Body map based wound image classification using deep learning(I.O.E. Pulchowk Campus, 2023-05) Khanal, BibekIdentifying different types of ulcer and surgical wounds based on their distinct features is a complex task in medical imaging. This involves the classification of ulcer and surgical wound into various labels such as diabetic ulcer, pressure ulcer, venous ulcer and surgical wounds. In order to make this process more efficient and cost-effective, there has been different study in this field. A body map based VGG 16 network is used to implement transfer learning onto two trainable dense layers for classification of wound images into five labels. The five labels include the aforementioned four types of wound and another label "Not a wound" which does not contain any wound image. The study is started with AZHMT dataset containing 4790 images. These images are classified using pre-trained inceptionV3 and VGG 16 network separately. The performance of VGG 16 was found to be better than inceptionV3 by almost 4% which was the reason for selecting VGG 16 for further study in this dataset. Also, inceptionV3 is longer and wider than VGG 16 which will learn unnecessary features from images using higher computing resources. The main aim of this thesis is to show that performance can be increased without learning unnecessary features, using fewer computing resources. and by using body map function.Item 5G COVERAGE PLANNING FOR URBAN AREA AT KATHMANDU CITY, NEPAL(IOE Pulchowk Campus, 2022-09) SHARMA, NIRMALAThe excitement about the 5G wireless network has passed. Mobile Network Operators (MNO) have begun rolling out 5G networks alongside 4G cellular networks in lower frequency and mid-frequency bands (i.e., 3-6 GHz) all over the world. The midfrequency band can greatly improve the performance of the current network (i.e., 50 MHz–100 MHz). All we know that the wider spectrum can be provided by the high frequency bands which is required to fulfill the greatest bitrates (20 Gb/s), lowest latencies, and constantly increasing capacity demands. The free space propagation loss rapidly increases as we move to higher frequency bands, which will reduce the individual cell site radius for the high-frequency band to 100 m from several kilometers in 4G. To offer consistent 5G coverage, the MNOs will have significant challenges in precisely planning and acquiring these enormous numbers of new cell site locations. This paper describes about the signal characteristics at 800MHz, 1800MHz for 4G and at 700MHz, 2300MHz, 2600MHz, 3500MHz for 5G and the upgradation of 4G towards 5G in the test environment. The 5G Coverage Planning with three sector cells and its SINR Mapping in advance antenna array will be performed to provide better coverage in 5G environments.Item INTERNET OF VEHICLE (IOV) BASED DRIVER EMOTION DETECTION USING FEDERATED LEARNING(IOE Pulchowk Campus, 2023-04) Regmi, Pankaj NidhiWith the advancement of edge smart computing devices and Internet of Vehicle (IoV) technologies emotion detection has become one of the most used methods in smart vehicle while driving. Many models have been employed however, privacy disclosure and communication cost are still a question. To address this question a federated learning driver emotion detection system model is proposed. It intelligently utilizes collaboration between edge, client and cloud for realizing dynamic model also protecting edge data privacy. Federated Learning has an advantage on privacy. In this thesis two different algorithm FedAvg and FedSGD are compared. It is found that accuracy of FedAvg is better than FedSGD. Also, FedSGD takes more steps to converge than FedAvg.Item Blockchain-Based E-Voting With Zero-Knowledge Proof Using Smart Contracts(IOE Pulchowk Campus, 2022-09) Alam, JunedThe data of the public block chain, being available to all nodes, it is necessary to hide the vote preference of the voter, and preserve the integrity of the casted vote, while at the same time, it is necessary to show that the voter has already voted, to prevent someone from casting multiple votes. This thesis work proposes an e-voting system using block chain and its smart contract as the rule setter. Here, with the help of the Paillier Cryptography system, the zero knowledge proof was accomplished. The zero knowledge proof here was used to show that the voter has already voted while at the same time, hiding the casted vote. The homo-morphic additive property of the Paillier cryptography system was used to perform addition on the encrypted cipher texts without the need to decrypt the cipher text to reveal the votes in the process. In the end, a secure voting mechanism was achieved.Item HYBRID QUANTUM-CLASSICAL DEEP LEARNING MODEL FOR PREDICTION OF COVID-19 USING CHEST X-RAY IMAGES(IOE Pulchowk Campus, 2022-09) BARAL, ANISHMedical images are difficult to collect and are full of insecurities and expensiveness.Pandemic such as COVID-19 break out suddenly and may be transferable from one person to another ,so we need to identity the victim and isolate them.Prescience of less datasets of such cases are difficult for the classical convolution model for prediction of disease.We need a high performance and accurate image classification model that assists doctor in diagnosis.The CNN layer of deep learning is also computationally complex as it need a lot of weights to train for better performance ,this increases the computational complexity of the model.Therefore,it is very necessary to develop a model which is fast,accurate and computationally efficient model.Here,we present a hybrid quantum classical convolution neural network for image classification.We run the model in simulators and different real quantum devices.We found that the hybrid model with less trainable parameters with low resolution and small training images was able to outperform the classical convolution neural network.The best hybrid quantum -classical model in this work was with accuracy of 0.9348 and 12318 trainable parameters.The best classical model was with accuracy of 0.9076.The computationally efficient model was with accuracy of 0.9239 with 2355 learn able parametersItem Adaptive Equalization of Fast-Time Varying Channel for EDGE System Enhanced by Dual Symbol Rate(Pulchowk Campus, 2013-11) Twanabasu, DineshOur modern society has transformed to an information-demanding system, seeking voice, video, and data in quantities with high mobility that could not be imagined even a decade ago. High data rate high mobility creates challenges to the communication system. The challenge to the system is to conceive highly reliable system unaffected by the problems caused in the multipath fading wireless channels. Broadband single carrier modulated signals experience severe multipath distortion scrambling & ISI when propagating through physical medium. To mitigate the effect of channel a transmission burst includes a training sequence which is known to the receiver and depending upon the effect of the channel in the received training sequence the equalizer updated. In mobile communication due to the relative motion between the mobile equipment and network element the channel experienced is time varying. The properties of the equalizer used for tracking such time varying channel have to be time varying hence adaptive. This thesis focuses on fast time varying channel where each radio burst experiences varying channel response. However for fast fading channel where each radio burst experiences varying channel response a single training block in a radio burst may not be sufficient to track the channel. In this thesis two training sequence are used in a single burst of a DSR EDGE. A slight modification is done in the burst structure to include two training sequence keeping the number of symbol in a frame constant. A least mean square error based algorithm is used for equalization of fast time-varying channel. The simulation of the system was done in Simulink. The simulation of the communication system showed the equalization and estimation of the symbols using two training sequence has better performance than that with single training sequence used in DSR EDGE frame.Item MITIGATION OF INTER CELL INTERFERENCE IN MULTI CELL ORTHOGONAL FREQUENCY DIVISION MULTIPLE ACCESS SYSTEMS(Pulchowk Campus, 2013-11) SHAKYA, SANUJOrthogonal Frequency Division Multiple Access (OFDMA) is the multi-user version of Orthogonal Frequency Division Multiplexing (OFDM). Due to the orthogonality of the sub-carriers, the interference between the sub-carriers is eliminated. However in multi-cell OFDMA system, if same sub-carriers are assigned to different users in neighboring cells, then inter-cell interference (ICI) occurs. ICI is more prominent for the users at cell boundaries due to which the cell edge users experience lower data rates compared to the users close to the base stations. During the thesis work, a dynamic radio resource allocation algorithm has been designed for mitigation of ICI. This algorithm assigns sub channels to the users and their transmission power based on the user location and sub channel assignment information from neighboring cells. The performance of Reuse 1, Reuse 3, Partial Frequency Reuse (PFR), Soft Frequency Reuse (SFR) and the dynamic resource allocation algorithm have been analyzed on the basis of change in Signal to Interference and Noise Ratio (SINR) values, system Bit Error Rate (BER), bandwidth efficiency and channel capacity with increase in distance of User Equipment (UE) from the center of cell. The dynamic resource allocation algorithm provides satisfactory levels of SINR values throughout the cell region and also maintains higher system capacity for users as compared to other resource allocation techniques.Item An Approach for Performance Enhancement of Secure Routing Protocols in Mobile Ad Hoc Networks(Pulchowk Campus, 2015-11) Timilsena, Santosh RajMobile ad hoc networks (MANETs) represent complex distributed systems that comprise wireless mobile nodes that can freely and dynamically self-organize into arbitrary and temporary network topologies, allowing people and devices to seamlessly internetwork in areas with no pre-existing communication infrastructure. Although the principle of wireless, structure-less and dynamic networks is attractive, there are still some major flaws that prevent its commercial expansion. Security is one of these main barriers. The open and dynamic operational environment of MANET makes it vulnerable to various network attacks. The security goals can be achieved using secure routing protocols. Ad hoc On-demand Distance Vector (AODV) is one of the most widely used routing protocols that is currently undergoing extensive research. This thesis presents the AODV protocol and surveys security enhancements using both cryptographic and trust based approaches. The impact of security features on routing performance was analyzed. The addition of a trust based approach with cryptographic features reduces routing overheads significantly. The proposed mechanism offers more resilience to attack from malicious nodes, while also promotes collaboration among cooperative nodes and penalizes selfish nodes. The simulation results show that the proposed trust model increases routing efficiency and reliability at cost of delay.Item NON LINEAR DISTORTION AND COMPENSATION TECHNIQUE IN WIMAX BROADBAND(Pulchowk Campus, 2015-11) ADHIKAREE, PRADIPWiMAX (Worldwide Interoperability for Microwave Access) is a wireless communications standard designed to provide 30 to 40 Mbps data rates. The physical layer of WiMAX is based on Orthogonal Frequency Division Multiplexing (OFDM) technique. OFDM is a Frequency Division Multiplexing (FDM) technique used as a digital multi-carrier modulation method. The major advantages of OFDM are robustness in multipath fading and high spectral utilization efficiency. However, it has some serious problem like highly sensitive to timing and frequency offset, and more specifically Peak-to-Average Power Ratio (PAPR) problem. Since OFDM is constructive superposition of the subcarriers, it has significant numbers of large peak resulting in high PAPR causing large fluctuation in input signal which requires the use of highly linear amplifier. Due to the amplifier imperfection the peaks are distorted non-linearly which generates inter modulation product causing both in-band distortion and Out-of-Band (OOB) radiation. The design of a compensation technique to improve the linearity of the power amplifier is performed in this thesis work. The adaptive pre-distortion method to compensate the non-linearity of power amplifier includes the LSE estimation of non-linearity and pre-distortion techniques. The analysis of Phase Realignment (PR) and Modified Phase Realignment (MPR) techniques for PAPR reduction for the performance improvement of the system showed that the combination of predistortion and PR/MPR technique has superior capability in mitigation of the nonlinear distortion as compared to the implementation of individual techniques.Item OPTIMIZATION OF ENERGY DETECTION APPROACH OF SPECTRUM SENSING IN COGNITIVE RADIO NETWORK(Pulchowk Campus, 2013-11) BARAL, NIRANJANCognitive Radio is the emerging technology that allows dynamic access of radio spectrum. Spectrum Sensing is the first task needed to be done to check the presence of licensed users in the spectrum. This thesis is focused on understanding the underlying principles of “Energy detection for Spectrum Sensing” in Cognitive Radio technology which does not need any prior information about the type of signal and optimizing its performance. In this research, spectrum sensing algorithms basically Energy Detection (ED) is considered under a typical fading unknown channel and White Gaussian Noise scenario. Knowledge of the noise power is imperative for the optimum performance of ED. Unfortunately the variation and unpredictability of noise power is unavoidable. Introducing an idea of auxiliary noise variance estimation for combating the absence of prior knowledge of noise power, Hybrid Energy Detection 1 (HED1)/Hybrid-2 (HED2) approach of signal detection was set forth. For HED noise variance is estimated in S auxiliary noise only slots and for HED2 noise variance is estimated in S auxiliary slots which are declared only noise signal slots by ED. The detection performance of the considered methods are derived and expressed by a closed form analytical formulas. The impact of noise estimation accuracy on the performance of ED is compared based on Receiver Operating Characteristic curves and Performance Curves. Accordingly, this study shows that even if the performance gap may be significant under some circumstances (few sensors, low signal-to-noise ratio, small number of slots used for noise power estimation), the performance gap can be decreased in terms of ROC performance by increasing the number of slots used for noise variance estimation.Item Performance Analysis of Space Time Trellis Code on Rayleigh, Rician and Nakagami Fading Channel(Pulchowk Campus, 2015-11) Khatiwada, Khem NathSpace-time codes merge coding gain with diversity gain and maintain orthogonality between the antennas. Coding works as a promising technique in the reliable digital data communication. STTC provides coding gain as well as diversity gain which gives the additional SNR advantage due to coding gain. This thesis analyzes the STTC over Rayleigh, Rician and Nakagami fading channel along with its implementation. STTC has been used for encoding and vitebri decoder for decoding. The design of 2-PSK for the SISO and use of 2 transmit antennas at transmitting side and use of receive antenna at receiving side up to 4 has been presented for 4, 8, 16 and 32 state. The result shows a significant improvement in performance of STTC with increasing number of states, number of transmit and receive antennas. It makes the data reliable and secured, along with increased throughput between the transmitter and receiver. It has been concluded that STTC over MIMO channel has improved error performance in terms of SNR with Nakagami fading channel.Item PERFORMANCE ANALYSIS OF COGNITIVE RADIO NETWORKS FOR RESOURCE SHARING(Pulchowk Campus, 2013-11) CHAPAGAIN, KAMALModern wireless applications demand the higher data rates and that led to the problem of scarcity of spectrum. According to Federal Communication Committee (FCC) 70% of statically allocated spectrums are underutilized and thus it had given legal permission for unlicensed operation for the Television White Space (TVWS). The concept of TVWS is due to the shifting of analog TV to Digital TV, traditionally reserved frequency bands in the range of 54 MHz to 698 MHz with 700 MHz band. Field surveying during this research work within the range of 0.1 MHz to 3 GHz also supports the FCC data because just about 17% of spectrum is used and apart from FM radio and Global System for Mobile (GSM) band, other licence band are almost void. In such a scenario Dynamic Spectrum Allocation (DSA) has got the higher priority and hence new concept of cognitive radio technology was emerged. Cognitive radio technology is an innovation software defined radio design that is proposed to increase the spectrum utilization by exploiting the unused spectrum with dynamically changing environment. This thesis envisions the current scenario of spectrum utilization and proposed a model for the secondary user which is not only an opportunistic user but also a co-operative user of primary user that assist for the transmission of data packet of primary user according to queuing theory. Thesis work focuses the DSA for Wireless Regional Area Network (WRAN)- IEEE 802.22, which is the first worldwide standard on cognitive radio because this offers ten times the coverage and three times better the capacity of current Wi-Fi spectrum. The hypothesis- performance is increased due to sharing of resource is became true by assuming the upper bound of 1 GB primary's data on queue and on sharing of about 50% of this data by the secondary transmitter as co-operative user is tested and found that the performance is increased by about 10% to 41% as traffic load increased upto 98%. Since, the data rate is set according to WRAN parameter as 24 Mbps, to get optimum performance of resource sharing is also achieved about 41% when traffic load is 0.98.Item DESIGN OF DIPLEXER USING WAVEGUIDE TECHNOLOGY(Pulchowk Campus, 2015-11) Tandukar, JatinThe rapid growth in wireless network has created a massive demand for concise frequency selective components. Diplexers are key components in varieties of communication systems. They are passive devices which implement frequency domain multiplexing and allow various frequencies to share common channel. A novel diplexer is a frequency selective device. It has three ports and six resonators where common junction is replaced by one of the resonator. The diplexer forms an all-resonator based structure and is designed and simulated in the CST Microwave Studio. The diplexer is designed using standard dimensions for rectangular cavity to operate in the X-band and it is based on the synthesis of coupling matrix of a three port coupled resonator. The completed design consists of three ports; one input port with centre frequency of 10 GHz and two output ports with centre frequencies 9.7 GHz and 10.3 GHz respectively with bandwidth approximately around 400MHz .The response of the designed diplexer is carefully studied and the response is good as compared to the response of the ideal diplexer.Item Performance analysis of Orthogonal-frequency-division-multiplexing (OFDM) based on Channel Estimation Technique(Pulchowk Campus, 2013-09) Bhandari, InduIn order to achieve the potential advantages of OFDM based systems, the channel coefficients should be estimated with minimum error. Orthogonal frequency division multiplexing (OFDM) is a method to transmit multi carrier in wireless environment, and can also be seen as a multi-carrier digital modulation or multicarrier digital multiplexing technology. Channel estimation plays a vital role in OFDM system. The channel estimation technique that can be pilot based or blind based can be helpful to improve the performance of OFDM system. The channel estimation technique for OFDM systems based on pilot arrangements is investigated. A different algorithm for both estimating channel at pilot frequencies and interpolating the channel is performed. The Performance comparison of all schemes by measuring bit error rate and mean square error using different modulation techniques like QPSK, and 16-QAM is discussed. Simulation results show the performance of different type channel estimation techniques under various channel condition and modulation technique.Item “Analysis and Prediction Of Sales Of The Company Using Data Mining”(Pulchowk Campus, 2012-11) Khadka, SumitThis thesis helps us to predict the future sales of the company on the basis of historical data and also to analyze the sales of the company. In this thesis different prediction or forecasting model is used to predict the future sales of the company and analyze the sales values given by the corresponding model and also, which prediction based model perform and predict better when no of training data increases and check dependencies of model with number of data sets in database and also which algorithm perform better on short term forecasting, mid term forecasting and long term forecasting on the basis of performance indices.