Loading...
Theses
Browse
13 results
Search Results
ThesisItem Open Access Multi-label classification of news titles using bidirectional long short-term memory model(G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2021-12) Goel, Yash Kumar; Samantaray, S.D.Multi-Label Text Classification can be used when there are two or more classes as well as the information to be classified may relate to neither of the classifications or all of them at the very same time. With the rapid advancement of devices and digital telecommunications, online news has become one of the most important attributes for people's daily lives, studies, and jobs. Online news, in comparison to other conventional media, is extensive, diversified in form, and could be updated in real-time. The lack of classification makes it hard for a person to interpret or obtain data relevant to particularly preferred classifications. Text classification, as one of the key technologies of information resource organization & management, could allow users to narrow the scope of feature extraction as well as make it more convenient as well as effective to filter via massive digital information to fulfil the needs. The technique of text classification, which in the classification stage is capable of classifying instantly against several classifications on unstructured text with natural language, is used. In the proposed work, WordNet and word sense database is used to improve the efficiency of the classifier. To handle a huge amount of data the classification deep learning approach i.e. Bidirectional Long/Short-Term Memory (Bi- LSTM) is proposed. As News Titles is a short text that could lead to ambiguity in classification class and the title of the news item could be linked to a number of different sources that could lead to ambiguity in classification class, the introduction of the phrase seeks to optimize the classification method. The challenge of news classification begins with web scraping to gather real-time news Titles from news websites, which are then instantly classified using different classification methodologies and introduce the Wordnet and WordSense database for multi-label news titles classification. The acquired accuracy of (Bi-LSTM) was 97.91 per cent, which exceeded the approximate accuracy of each individual plan. This technique could be very helpful for academicians who want to investigate headlines in order to support their instruction.ThesisItem Open Access Stock price prediction using LSTM approach(G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2021-09) Agarwal, Shweta; Singh, B.K.In today's economy, the stock market, often known as the equity market, has a significant impact. The rise or decline in the share price has a significant impact on the investor's profit. The proposed method used Long short-Term Memory (LSTM) Approach. Here I am considering multi-column LSTM model which takes more than one column to analyse and train the model and based on that it will predict the values for future days. More than one features helps the model to predict the values more accurately than providing the single feature. Here the dataset is taken from Yahoo Finance website which provides historical data to almost all of the companies listed in the stock market. The dataset is taken for a particular company PETRONET LNG from 2004 to 2018. Next 30 days values are being predicted based on that historical data. The values for 2019 is not being considered as this time was affected by corona virus and every sector of the industry was affected by this pandemic. So taking these values may provide wrong predictions as there was sudden fall and rise in the stock values during this time. I have also added 2 more features to the given historical data i.e. volatility and momentum. Volatility is basically used to capture fluctuation in the market. Momentum tells us what is the changes in the price as compare to past days. Result showa that adding these features helps model to predict more accurately.ThesisItem Open Access Cloud based real time soil moisture content monitoring using IOT and unmanned aerial vehicles(G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2021-09) Joshi, Kritika; Singh, RajeevRecently, agricultural UAVs or drones have become one of the most useful agricultural instruments utilized in smart farming, especially in the ground sensing applications. The UAV is employed as a mobility hotspot in the area lacking of the communication network to collect the data of soil and crops in order to monitor and control the water management system for precision agriculture. The agricultural UAV with a combination of IoT sensors will be effected in such areas. At present, the IoT sensors detect temperature and moisture of the soil and send them to farmers by using the cloud but not by using UAV. In this work we combined and connected the UAVs and IoT network based sensor. By this connection, the identification of locations that suffer from droughts, water scarcity, and dryness of soil profile can be easily observed. It will help the farmers to take precautionary methods such as contouring, damming, draining the surface water, and curtailing for further irrigation. The objective of this work is to design a system to collect and measure the soil moisture contents remotely in real-time through drone. For a country like India whose main source of revenue for citizens is agriculture, this domain is not explored sufficiently in terms of technology. The lack of sufficient research in this domain has been a problem for the livelihood of those farmers whose sole existence depends on agriculture. An economical process could be devised that could help automate the process of knowing the exact soil moisture content of the entire farmland. Such process would be a boon for the farmers. above work is only a step forward in this direction. In this thesis work we monitor real time soil moisture by collecting the data using the soil moisture sensor and Arduino UNO. The transmitted data is transmitted to the drone who further send the resultant data for real time monitoring and analysis to the cloud.ThesisItem Open Access Evaluation of controller placement approaches in SDN-based 5G networks(G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2021-09) Agarwal, Lakshita; Sharma, JalajThe thesis work provided the relevant information about the SDN-based 5G networks, with the controller placement problem as well as the average latency between the nodes of the networking system. Software-defined networking (SDN) is a technology that is emerging in today’s world and it is mainly used for changing the state of the network by breaking the older version of it and by separating the network’s control layer from the layer of the routers and switches. With the evolution of the fifth generation, i.e., 5G networks, there is a steady growth in the development of different business models as well as new applications are being developed. For the development of the 5G networking and for the formation of different intelligent networks and applications, SDN technology has been considered as a key enabler. The main aim of this work is about the evaluation of the already existing controller placement approaches i.e., K-Median, K-Center and Kcritical Approaches for SDN-Based 5G networking system. The proposed work is an approach for analyzing the solutions for the placement of the controller inside the SDNbased 5G network. In this work, the network topology of 250 nodes was created and the evaluation of the three different approaches was being done. It was also concluded that out of all the three approaches the K-Critical approach can be considered as one of the best approaches for the selection of the controller inside the SDN-based 5G networks because all the criterions in this approach were satisfied according to the requirements.ThesisItem Open Access Using ensemble and TOPSIS with AHP for classification and selection of web services(G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2021-09) Pandey, Mithilesh; Jalal, SunitaDue to the increasing number of web services with the same functionality, selecting a web service that best serves the needs of the Web Client has become a tremendously challenging task. Present approaches use non-functional parameters of the web Services, but they do not consider any preprocessing of the set of functionally similar web services. Due to the lack of preprocessing, the web services selection method also has to process web services with a very low to no chance of satisfying the consumer's requirements. This thesis proposes an Ensemble classification method for preprocessing and a web services selection method based on the Quality of Service (QoS) parameters. Once the most eligible web services are enumerated through classification, they are ranked using the TOPSIS (Technique of Order Preference by Similarity to Ideal Solution) method with AHP (Analytic Hierarchy Process) used for weight calculation. A prototype of the method is developed, and experiments are conducted on a real world web services dataset. Results demonstrate the feasibility of the proposed method.ThesisItem Open Access A hybrid Artificial Bee Colony Genetic Algorithm (ABCGA) approach for energy efficiency in wireless sensor network communication(G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2021-09) Pandey, Garima; Mishra, P.K.Non-rechargeable batteries with restrained energy capability are used to power the wireless sensor network nodes, but in hostile conditions, replacing a node battery is a difficult task. Thus, enhancing the energy efficiency resulting in increasing the lifetime of the network is a suitable decision. Recent work suggests that clustering is an efficient mechanism for reducing energy consumption, increasing network scalability, maintain load balancing, all of which contribute to maximizing total network life. Appropriate cluster head selection in a cluster is crucial as it prominently affects the wireless sensor network life. Metaheuristic algorithms can be utilised effectively for this.This thesis main contribution is to design a bee colony optimization technique named ABCGA dependent on an artificial bee method. In the proposed algorithm ABCGA, theartificial bee method is integrated with the features of the genetic algorithm for optimal cluster head selection. Also, data compression is done before data transmission, resulting in reduced energy consumption and increased network life. This proposed method has been evaluated against the LEACH, PBC-CP, PSO, and HSAPSO techniques in terms of the following factors- the number of active nodes, the number of nodes dead, remaining energy, and throughput. In the end, the simulation findings demonstrated that the proposed approach ABCGA outperforms all these four standard algorithms.ThesisItem Open Access Identification of nutritionally important protein in Amaranthus genes using sequence mining(G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2021-02) Dheeraj Kumar; Samantaray, S.D.Various herbs have been used since ancient time to solve physical problems. It is known that a traditional discovery of Amaranthus plant can neutralize unique diseases. In addition, it can help to overcome the system of disease and increase the effect of scientific treatment and medicine. Amaranthus can be considered a safe haven for wellbeing, due to its therapeutic properties. It affects severe physical problems involving coronary disease, malignant growth, inflammation of joints, stagnation, and liver-kidney problems. As the world's population is increasing day by day and ground, water and food resources are limited, it is of the utmost importance that good sources of protein should be included in human diets, keeping in mind the quantity and quality of proteins required to meet human diets. This article gives a comprehensive idea of Amaranthus that focuses on the research reporting its use in the medical trials and all of its profit to human health. The purpose of this research is to detect the presence of nutritionally important protein sequences in Amaranthus Genes using BLAST (tBLASTn), classification techniques and sequence mining techniques. We found the high similarity searches for Ama1 protein in Amaranthus genes having the following results: maximum bit score 304, total bit score 1247.2, gap 0/146 (0%) and E-Value (min) 4e-134. The second largest similarity found for Dreb1a protein having the following values: maximum bit score 192, total bit score 435, gap 10/197(5%) and E-Value (min) 2e-56.ThesisItem Open Access Use of drone in disease identification from leaves by deep learning through YOLO v3 and CNN architecture(G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2021-02) Neha; Singh, RajeevIn recent years Machine Learning and Deep Learning have played a crucial role in the field of agriculture. There are many methods that are adopted in farming so that the yield and production increases. Smart agriculture is itself growing and developing. In automated farming, smart agriculture helps to collect data from the field and then analyze it so that the farmer can make precise decisions to grow high-quality crops. For better agricultural productivity and food management, an agriculture monitoring system is needed. Precision agriculture is also used as new technology for the decision making process. In this work, we have used drone for collecting data in real time from the field. ML algorithm are then used to take optimal decisions which helps in cutting the cost of procedure. Drone systems are also used reliably for operations like UREA spraying wherein involvement of the sensors enables a reliable safe operation with good satisfaction of customer. However, this field is open for improvements majorly in decision support system which helps in converting large amount of data into useful recommendations. Deep learning is a subset of machine learning. It can be used for precision farming, identification of diseases, classification of images etc. This research deals with identification of wheat plant leaf diseases by accessing the leaf morphology of crops by means of drone photography and further analysis of captured images by computer means using YOLO V3 and CNN architecture.ThesisItem Open Access Music genre classification using RNN-LSTM approach(G.B. Pant University of Agriculture and Technology, Pantnagar - 263145 (Uttarakhand), 2021-02) Lokesh Kumar; Singh, B.K.In the present world of technology, computers and computational techniques play a very keen role in the life of each individual. Everyone is dependent on technology in some or another way, whether it be personal or professional dependability. In the era of Artificial Intelligence and Machine Learning, complex real-world problems are being solved by these evolving techniques. Now a days, the difficulties and cost of biological analyses have led to the development of sophisticated machine learning approaches. In this work, a self constructed dataset that consists of 6 classes with each class representing a sub-genre namely Abhang, Bhajan, Kajari, Qawwali, Tappa, and Thumri, that fall under the Indian semi-classical genre hierarchy. In this research work the above mentioned classes of music has been classified. The RNN-LSTM (Long-Short Term Memory) deep learning technique has been used here for classification. We have used STFT (Short-Term Fourier Transform) and DCT (Discrete Cosine Transform) to pre-process the music dataset. This research can be helpful in developing an automatic music recommendation module of online music applications, for increasing the browsing functionality of the music platform.
