http://hdl.handle.net/10366/148479 Sat, 09 May 2026 06:15:08 GMT 2026-05-09T06:15:08Z http://hdl.handle.net/10366/148644 The subject of human activity recognition is considered an important goal in the domain of computer vision from the beginning of its development and has reached new levels. It is also thought of as a simple procedure. Problems arise in fast-moving and advanced scenes, and the numerical analysis of artificial intelligence (AI) through activity prediction mistreatment increased the attention of researchers to study. Having decent methodological and content related variations, several datasets were created to address the evaluation of these ways. Human activities play an important role but with challenging characteristic in various fields. Many applications exist in this field, such as smart home, helpful AI, HCI (Human-Computer Interaction), advancements in protection in applications such as transportation, education, security, and medication management, including falling or helping elderly in medical drug consumption. The positive impact of deep learning techniques on many vision applications leads to deploying these ways in video processing. Analysis of human behavior activities involves major challenges when human presence is concerned. One individual can be represented in multiple video sequences through skeleton, motion and/or abstract characteristics. This work aims to address human presence by combining many options and utilizing a new RNN structure for activities. The paper focuses on recent advances in machine learning-assisted action recognition./nExisting modern techniques for the recognition of actions and prediction similarly because the future scope for the analysis is mentioned accuracy within the review paper.; La temo de homa agado-rekono estas konsiderata grava celo en la regado de komputila vizio ekde la komenco de ?ia disvolvi?o kaj atingis novajn nivelojn. ?i anka? estas pensata kiel simpla procedo. Problemoj ekestas en rapidaj kaj progresintaj scenoj, kaj la nombra analizo de artefarita inteligenteco (AI) per agado-anta?diro mistraktado pliigis la atenton de esploristoj por studi. Havante decajn metodikajn kaj enhavajn rilatajn varia?ojn, pluraj datenserioj estis kreitaj por trakti la taksadon de ?i tiuj manieroj. Homaj agadoj ludas gravan rolon sed kun malfacila karakteriza?o en diversaj kampoj. Multaj aplikoj ekzistas en ?i tiu kampo, kiel inteligenta hejmo, helpema AI, HCI (Homa-Komputila Interagado), progresoj en protekto en aplikoj kiel transportado, edukado, sekureco kaj administrado de medikamentoj, inkluzive faladon a? helpon al maljunuloj pri kuracado de drogoj. La pozitiva efiko de profundaj lernaj teknikoj sur multaj vidaj aplikoj kondukas al disfaldi ?i tiujn manierojn en video-prilaborado. Analizo de homaj kondutagadoj implikas gravajn defiojn kiam homa ?eesto temas. Unu individuo povas esti reprezentita en multoblaj videosekvencoj tra skeleto, movi?o kaj / a? abstraktaj karakteriza?oj. ?i tiu verko celas trakti homan ?eeston kombinante multajn eblojn kaj uzante novan RNN-strukturon por agadoj. La papero temigas lastatempajn progresojn en ma?inlernado-helpata agado.Ekzistantaj modernaj teknikoj por la rekono de agoj kaj prognozo simile ?ar la estonta amplekso por la analizo estas menciita precizeco ene de la recenzo-papero. Tue, 08 Feb 2022 00:00:00 GMT http://hdl.handle.net/10366/148644 2022-02-08T00:00:00Z http://hdl.handle.net/10366/148645 Epilepsy is one of the most occurring neurological diseases. The main characteristic of this disease is a frequent seizure, which is an electrical imbalance in the brain. It is generally accompanied by shaking of body parts and even leads (fainting). In the past few years, many treatments have come up. These mainly involve the use of anti-seizure drugs for controlling seizures. But in 70% of cases, these drugs are not effective, and surgery is the only solution when the condition worsens. So patients need to take care of themselves while having a seizure and be safe. Wearable electroencephalogram (EEG) devices have come up with the development in medical science and technology. These devices help in the analysis of brain electrical activities. EEG helps in locating the affected cortical region. The most important is that it can predict any seizure in advance on-site. This has resulted in a sudden increase in demand for effective and efficient seizure prediction and diagnosis systems. A novel approach to epileptic seizure prediction and diagnosis system "EpilNet" is proposed in the present paper. It is a one-dimensional (1D) convolution neural network. EpilNet gives the testing accuracy of 79.13% for five classes, leading to a significant increase of about 6-7% compared to related works. The developed Web API helps in bringing EpilNet into practical use. Thus, it is an integrated system for both patients and doctors. The system will help patients prevent injury or accidents and increase the efficiency of the treatment process by doctors in the hospitals. Tue, 08 Feb 2022 00:00:00 GMT http://hdl.handle.net/10366/148645 2022-02-08T00:00:00Z http://hdl.handle.net/10366/148646 /n This work proposes a technique of breast cancer detection from mammogram images. It is a multistage process which classifies the mammogram images into benign or malignant category. During preprocessing, images of Mammographic Image Analysis Society (MIAS) database are passed through a couple of filters for noise removal, thresholding and cropping techniques to extract the region of interest, followed by augmentation process on database to enhance its size. Features from Deep Convolution Neural Network (DCNN) are merged with texture features to form final feature vector. Using transfer learning, deep features are extracted from a modified DCNN, whose training is performed on 69% of randomly selected images of database from both categories. Features of Grey Level Co-Occurrence Matrix (GLCM) and Local Binary Pattern (LBP) are merged to form texture features. Mean and variance of four parameters (contrast, correlation, homogeneity and entropy) of GLCM are computed in four angular directions, at ten distances. Ensemble Boosted Tree classifier using five-fold cross-validation mode, achieved an accuracy, sensitivity, specificity of 98.8%, 100% and 92.55% respectively on this feature vector. Tue, 08 Feb 2022 00:00:00 GMT http://hdl.handle.net/10366/148646 2022-02-08T00:00:00Z http://hdl.handle.net/10366/148641 Clustering is the unsupervised machine learning process that group data objects into clusters such that objects within the same cluster are highly similar to one another. Every day the quantity of Urdu text is increasing at a high speed on the internet. Grouping Urdu news manually is almost impossible, and there is an utmost need to device a mechanism which cluster Urdu news documents based on their similarity. Clustering Urdu news documents with accuracy is a research issue and it can be solved by using similarity techniques i.e., Jaccard and Dice coefficient, and clustering k-mean algorithm. In this research, the Jaccard and Dice coefficient has been used to find the similarity score of Urdu News documents in python programming language. For the purpose of clustering, the similarity results have been loaded to Waikato Environment for Knowledge Analysis (WEKA), by using k-mean algorithm the Urdu news documents have been clustered into five clusters. The obtained cluster's results were evaluated in terms of Accuracy and Mean Square Error (MSE). The Accuracy and MSE of Jaccard was 85% and 44.4%, while the Accuracy and MSE of Dice coefficient was 87% and 35.76%. The experimental result shows that Dice coefficient is better as compared to Jaccard similarity on the basis of Accuracy and MSE. Tue, 08 Feb 2022 00:00:00 GMT http://hdl.handle.net/10366/148641 2022-02-08T00:00:00Z http://hdl.handle.net/10366/148642 The fast growth of Internet and social media has resulted in a significant quantity of texts based review that is posted on the platforms like social media. In the age of social media, analyzing the emotional context of comments using machine learning technology helps in understanding of QoS for any product or service. Analysis and classification of user's review helps in improving the QoS (Quality of Services). Machine Learning techniques have evolved as a great tool for performing sentiment analysis of user's. In contrast to traditional classification models. Bidirectional Long Short-Term Memory (BiLSTM) has obtained substantial outcomes and Convolution Neural Network (CNN) has shown promising outcomes in sentiment classification. CNN can successfully retrieve local information by utilizing convolutions and pooling layers. BiLSTM employs dual LSTM orientations for increasing the background knowledge accessible to deep learning based models. The hybrid model proposed here is to utilize the advantages of these two deep learning based models. Tweets of users for reviews of Indian Railway Services have been used as data source for analysis and classification. Keras Embedding technique is used as input source to the proposed hybrid model. The proposed model receives inputs and generates features with lower dimensions which generate a classification result. The performance of proposed hybrid model was compared using Keras and Word2Vec and observed effective improvement in the response of the proposed model with an accuracy of 95.19%. Tue, 08 Feb 2022 00:00:00 GMT http://hdl.handle.net/10366/148642 2022-02-08T00:00:00Z http://hdl.handle.net/10366/148643 Flying Ad-hoc Network (FANET) is a special member/class of Mobile Ad-hoc Network (MANET) in which the movable nodes are known as by the name of Unmanned Aerial Vehicles (UAVs) that are operated from a long remote distance in which there is no human personnel involved. It is an ad-hoc network in which the UAVs can more in 3D ways simultaneously in the air without any onboard pilot. In other words, this is a pilot free ad-hoc network also known as Unmanned Aerial System (UAS) and the component introduced for such a system is known as UAV. There are many single UAV applications but using multiple UAVs system cooperating can be helpful in many ways in the field of wireless communication. Deployments of these small UAVs are quick and flexible which overcome the limitation of traditional ad hoc networks. FANETs differ from other kinds of ad hoc networks and envisioned to play an important role where infrastructure operations are not available and assigned tasks are too dull, dirty, or dangerous for humans. Moreover, setting up to bolster the range and performance of small UAV in ad hoc network lead to emergent evolution with its high stability, quick deployment, and ease-of-use for the formation of the network. Routing and task allocation are the challenging research areas of the network with ad hoc nodes. The paper overview based on the study of biological inspired routing protocols (Moth-and-Ant and Bee Ad-Hoc) routing protocols. Tue, 08 Feb 2022 00:00:00 GMT http://hdl.handle.net/10366/148643 2022-02-08T00:00:00Z http://hdl.handle.net/10366/148638 Tue, 08 Feb 2022 00:00:00 GMT http://hdl.handle.net/10366/148638 2022-02-08T00:00:00Z http://hdl.handle.net/10366/148640 The concept of data mining is to classify and analyze the given data and to examine it clearly understandable and discoverable for the learners and researchers. The different types of classifiers are there exist to classify a data accordingly for the best and accurate results. Taking a primary data, and then classifying it into different portions of parts, then to analyze and remove any ambiguities from it and finally make it possible for understanding. With this process, that data will become secondary from primary and will called information. So, the classifiers are doing the same strategy for the solution and accuracy of the data. In this paper, different data mining approaches have been used by applying different classifiers on the taken data set. The data-set consists of 500 candidates' segregated data for the analysis and evaluation to perfectly classify and to show the accurate results by using the proposed Algorithms. The data mining approaches have been used in which HUGO (Highly Undetectable steGO) Algorithm, Naïve Bayes Classification, k-nearest neighbors and Logistic Regression are used with the extension of the other classification methods that are Support Vector Machine (SVM) and Multi-Layer Perceptron (MLP) as classifiers. These classifiers are given names for further analysis that are Classifier-1 and Classifier-2 respectively. Along with these, a tool is used named WEKA (Waikato Environment for Knowledge Analysis) for the analysis of the classifier-1 and 2. For performance evaluation and analysis the parameters are used for best classification that which classifier has given best performance and why. These parameters are RRSE (Root Relative Square Error), RAE (Relative Absolute Error), MAE (Mean Absolute Error), and RMSE (Root Mean Square Error). For the best and outstanding accuracy of the proposed work, these parameters have been tested under the simulation environment along with the incorrect, correct classifying and the %age has been witnessed and calculated. From simulation results based on RRSE, RAE, MAE and RMSE, it has been shown that classifier-1 has given outstanding performance among the others and has been placed in highest priority. Tue, 08 Feb 2022 00:00:00 GMT http://hdl.handle.net/10366/148640 2022-02-08T00:00:00Z http://hdl.handle.net/10366/148639 Tue, 08 Feb 2022 00:00:00 GMT http://hdl.handle.net/10366/148639 2022-02-08T00:00:00Z