Thumbnail Image

Theses

Browse

2016 - 201822
Reset filters
Subject: Computer Engineering × End date: 2019 × Start date: 2016 × Type: Thesis × Thesis Type: M.Tech ×
Reset filters

Search Results

Now showing 1 - 9 of 22
  • Thumbnail Image
    ThesisItemOpen Access
    Performance evaluation of fuzzy C mean and simulated annealing based clustering in WSN
    (G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2018-08) Awasthi, Shivangi; Mishra, P.K.
    Cluster based routing technique is most popular routing technique in Wireless Sensor Networks (WSNs). Due to varying need of WSN applications, efficient energy utilization in routing protocols is still a potential area of research. In this research work, focus is made on the optimization of clustering and to balance the load over the routes so that energy can be used effectively. The proposed work tried to overcome the problem of random distribution of clusters in LEACH. In this study, the optimization of clusters is made by using Fuzzy C Mean Clustering that gives uniformity in the cluster due to central tendency so that a uniform density can be seen in the participating clusters. To further distribute the load of transmitting data through a planned routing scheme Simulated Annealing is used. The performance of proposed work is evaluated on MATLAB by comparing it with some existing protocols.
  • Thumbnail Image
    ThesisItemOpen Access
    Application of FCM, GLCM & DWT and SVM for disease identification in mango
    (G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2018-08) Rohan, J.; Singh, Rajeev
    India is basically known as agricultural nation where 70% of individuals rely on agribusiness, in this way plants play an influential factor in their life and furthermore assume an inescapable part in the ecological balance of the nation. With regards to plant sickness, there are numerous sorts of ailment existed in world contrasted from area to district. The plant ailments happen every now and again and contrast from each other. These sicknesses can lessen the nature of rural items and caused substantial misfortunes even undermined the sustenance security and for the most part caused irresistible creatures or different other ecological factor. At some point the plant illness taints other piece of plants like leaves or branches and prompts finish collect misfortune and even uphold sustenance shortage. A computerized acknowledgment and characterization framework for these rural items can upgrade its quality by perceiving sickness side effects prior and analyze it. Because of quick advancement in data innovation, it assumes a critical part in preparing, perceiving and characterizing the plant ailments. In this paper we proposed a picture based malady arrangement for mango natural product utilizing GLCM & DWT, FCM and SVM. GLCM and DWT are used for highlight extraction; FCM is utilized for segmenting the images and Support Vector Machine is used for classification.
  • Thumbnail Image
    ThesisItemOpen Access
    Fake news detection using text similarity approach
    (G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2018-08) Jodhani, Geetika; Samantaray, S.D.
    The present work proposes a methodology for detecting Fake News in online news media using Text Similarity Approach. In this era of digitization, most of the people now get news from internet and often it can be difficult to tell whether stories are credible or not. Information overload and a general lack of understanding about how the internet works by people has also contributed to an increase in fake news or hoax stories. Traditionally we got our news from trusted sources, journalists and media outlets that are required to follow strict codes of practice. However, the internet has enabled a completely new way to publish, share and consume information and news with very little regulation or editorial standards. The present work is aimed to develop an automatic fake news detection system for analysing the credibility of online news. So that the reader become aware about the news that is factually incorrect and optimized for sharing. News articles are nothing but a piece of text. Hence, the proposed work can be divided into two subtasks; Text Analysis and Performance Evaluation. Text analysis is done for the transformation of text into numerical features. These numerical features are then used for matching the similarity between queried article and other articles. For articles similarity I have used hybrid of three text similarity approaches, two methods from lexical similarity features (N-grams (Character Based) and Cosine Similarity method (Corpus Based) and one from semantic similarity feature (Explicit Semantic Analysis (ESA) - TF*IDF (Term Based Similarity)). Python 3.5 is used for programming. System is tested for 100 news articles and analysed that if more than three articles have matching with matching value ≥ 0.70 and < 0.80, then it will result to truthiness of the input article. Our proposed system has gained the accuracy of 91.67%.
  • Thumbnail Image
    ThesisItemOpen Access
    Classification of liver disorder from serum profile using extreme learning machine
    (G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2018-07) Kharayat, Shweta; Singh, B.K.
    A problem is said to be well posed if it fulfilled the following properties: 1) a solution exists, 2) the solution is unique, and 3) non-perturbation. There exists a special case of the well posed problem where the inverse of the major function does not exist. The prediction of liver disorder in human from serum profile is one such problem. There are several diseases that could possibly affect the liver, with ample category of symptoms. In liver along with symptoms variation, intensity and severity differ from nearly insignificant to life-threatening. The proposed work predicts liver disorder in humans by finding relationship between serum profile parameters and occurrence of liver disorder using machine learning approach. Extreme Learning Machine (ELM) has enough capabilities to solve this type of problem. Extreme Learning Machine approach is utilized for implementing the proposed work by predicting liver disorder in human using serum profile. The serum profile considered in the research work is Tb, Alkphos, Sgpt, Sgot and others. Model classifies on the basis of disease severity. The datset utilized in the thesis is compiled from the UCI machine learning repository. Model is evaluated on several samples of dataset utilizing confusion matrixes and ROC curves for different activation functions. Based on that accuracy of the model evaluation is performed and RBF found to be best activation function to be used in the liver disorder problem. Proposed model has an accuracy of 80.9% for Radial Basis Function.
  • Thumbnail Image
    ThesisItemOpen Access
    An ensemble based classification approach for credibility analysis of online news by detecting clickbait news headlines
    (G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2017-08) Agarwal, Parul; Samantaray, S.D.
    The present work proposes a methodology for detecting clickbait news headlines in online news media using Ensemble based classification Technique. In this era of Digitization, presenting news now became online. Everyone is accessing online news by one or other medium. When online news is so popular and easily accessible, it also makes online news vulnerable too. Anyone can write anything in the name of news and it becomes viral whether it is informative or not. Due to the high competition and thrust of clicks, clickbait headlines are manufactured just to attract readers to click. These headlines generate enough curiosity by using some tactics so that readers compelled to click on the link to fill the knowledge gap. Clickbait headlines are compromising the meaning of true journalism. The present work is aimed to develop a clickbait detection system for analyzing the credibility of online news. So that the readers become aware and do not click on these links. News headlines are a piece of text, hence the proposed task is divided into two subtasks; Text analysis and classification. Text analysis is done for the transformation of text into numerical features usable for machine learning. These numerical features are then used for training the ensemble based classifier. The training dataset contains 10000 clickbait and 10000 non-clickbait headlines. Python 2.7 is used for the programming and system is tested for 10600 news headlines which are in an even distribution of 5800 clickbait and nonclickbait headlines and gained 93.13% accuracy. This system is also validated using k-fold cross validation technique.
  • Thumbnail Image
    ThesisItemOpen Access
    Classification of Glaucoma and Bright Lesions in Retinal Fundus Images using SVM
    (G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2017-08) Upadhyaya, Himanshu; Negi, Chetan Singh
    In terms of population, India stands at the second position in the world, and with such a huge population it is very difficult to provide medical assistance to each and every one especially to those living in the remote areas. As these retinal diseases require regular check-ups and timely intervention to control the progress of disease, an ophthalmologist with all the medical equipment is required which could be highly expansive. Therefore an automated clinical support system should be developed for the diagnosis of retinal diseases like glaucoma and diabetic retinopathy which could be used to make the screening of real time population easy and efficient and also identify those who are at risk in the early stages. This technique would minimize the cost, estimation time and also assist the ophthalmologist to perform the treatment plan. This thesis presents a classification system for the diagnosis of Glaucoma and Bright Lesions in retinal fundus images where different anatomical and statistical features are extracted and classified using SVM. It has been observed that the anatomical features proved to be a promising features as compared to the other statistical features and a good accuracy is achieved using SVM classification. In this thesis work the performance analysis of this classification system over different feature sets is reported and discussed.
  • Thumbnail Image
    ThesisItemOpen Access
    Encoder-decoder based integrity verification for video surveillance
    (G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2017-08) Panwar, Amit; Singh, Rajeev
    Surveillance cameras are widely used anywhere to record video data. In order to counter the editing or copying by adversaries, assuring the integrity of the extracted video is one of the fundamental issues in this area. Video surveillance is increasing significance as organizations seek to safe guard physical and capital assets. At the same time, the necessity to observe more people, places, and things coupled with a desire to pull out more useful information from video data is motivating new demands for scalability, capabilities, and capacity. Two improved system are described for verifying video content integrity, one uses frame level integrity and other uses digital watermarking. Existing verification systems are unable to distinguish between attacks and regular modifications and are thus unsuitable countermeasures against actual threats. The first proposed method helps in identifying the distortion in the video data at the frame level. The second proposed method distinguishes attacks against video content from regular modifications by extracting time codes and header hash values embedded in the content itself and comparing them with the actual ones, making it well suited for content storage services. Evaluation showed that second method is more effective than the one using the digital signature scheme.
  • Thumbnail Image
    ThesisItemOpen Access
    A novel approach for mapping of a boolean function using artificial neural network
    (G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2017-07) Sharma, Jalaj; Samantaray, S.D.
    Boolean functions have enormous importance in the field of Computer Engineering. These functions are important not only because of the fact that computer hardware architecture is based on them but also because of ever-increasing automation using devices like programmable logical controllers that utilize such functions in ladder programming. Another aspect of the importance of Boolean functions is the ability to transform any number to binary form for the purpose of any kind of processing or analysis of the data. Owing to this fact, various methods for representation of Boolean functions have been suggested in the literature. The input-output relationship for the devices based on Boolean functions can be mapped using trained artificial neural networks. Artificial Neural Network (ANN) is an intelligent tool with parallel computational capability. Conventional ANN once designed needs to be trained using iterative training process. The proposed method for mapping of Boolean functions is advantageous because of its generality and ease of implementation. This method is based on a novel neural architecture known as Pi-Sigma neuron model. Although Pi-Sigma neuron model is complex but the proposed new Pi-Sigma neuron model named as Simplified Pi-Sigma neuron model reduces the complexity and makes the learning process simple and non-iterative. While the conventional neuron models have summation operation for aggregation, the proposed neuron model has multiplication as well as summation operations for representing aggregation of the dendritic inputs. Incorporation of the multiplication operation along with the summation operation is based on some biological evidences as observed by researchers in the field of computational neuroscience. As any Boolean function can be represented in terms of sum of products, the proposed neuron model is capable of representing any Boolean function because of its inherent nature of performing multiplication operations before performing summation operations. The advantage with this method is that it does not require a long process of iterations for training. Weights and biases are directly calculated by presenting the training data in a single stroke. The proposed model works primarily for Boolean functions, but it can be extended to any kind of functions by using the conversion of number systems along with this method of functional mapping.
  • Thumbnail Image
    ThesisItemOpen Access
    Threshold sensitive energy efficient multi-sink routing protocol for heterogeneous wireless sensor networks
    (G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2017-08) Sumit Kumar; Mishra, P. K.
    The prominence of Wireless Sensor Networks (WSN) has expanded massively in late time because of development in Micro-Electro-Mechanical Systems (MEMS) innovation. WSN has the possibility to associate the physical world with the virtual world by framing a system of sensor nodes. Here, sensor nodes are typically battery-driven gadgets, and consequently energy consumption of sensor nodes is a noteworthy design issue. This thesis addresses ‘WSN’s lifetime and stability period optimization problem” which is to design an energy efficient protocol in such a way that energy consumption of every node in the wireless sensor networks is minimized which results in an improved stability period and prolonged lifetime of WSNs. This thesis solves the problem by introducing a static clustering based multi-sink routing protocol for heterogeneous wireless sensor networks. The idea of threshold aware transmission is also utilized to accomplish these objectives. The results are compared with two well known traditional clustering protocols namely LEACH and SEP using stability period, network lifetime, instability period and throughput as performance metrics. The proposed work performs better than the other protocols under consideration.