Background Stroke is the leading worldwide cause of disability and death. Effective stroke prevention and management depend on early identification of stroke risk. Methods Eight machine learning algorithms are applied to predict stroke risk using a well-curated dataset with pertinent clinical information. This paper describes a thorough investigation of stroke prediction using various machine learning methods. Results The empirical evaluation yields encouraging results, with the logistic regression, support vector machine, and K-nearest neighbors models achieving an impressive accuracy of 95.04%, and the random forest and neural network models scoring even better, with accuracies of 95.10% and 95.16%, respectively. The neural network exhibits slightly superior performance, indicating its potential as a reliable model for stroke risk assessment. Conclusions The empirical evaluation underscores the ability of neural networks to discern intricate data relationships. These findings offer valuable insights for healthcare professionals and researchers, aiding in the development of improved stroke prevention strategies and timely interventions, ultimately enhancing patient outcomes.

Image classification is one of the most significant applications of Deep Learning models. Deep Learning forms a subset of machine learning wherein the neural networks consist of more than three layers. There are various popular deep-learning approaches for image classification. In this paper, we have analyzed deep learning models and activation functions to compare their efficiency and accuracy. The dataset used consists of 15 classes of vegetables each having 1000 images for training, 200 images for validation, and 200 images for testing. Types of models analyzed here include Multi-layer Perceptron Model, CNN Model, and Pre-trained model. The different activation functions used are ReLU, Leaky ReLU, ELU, SELU, Sigmoid, and Tanh. The results of the empirical evaluation for image classification yield that a CNN model is better than a perceptron model as it achieved high accuracy and low loss in less iterations when compared with the perceptron model. Relu activation function yields maximum accuracy when used with CNN for image classification.

The novel coronavirus (COVID-19) has devastated millions of people and is a major threat to world health. The world economy was severely disrupted, millions of people died, and many suffered from severe psychological illnesses. Therefore, by employing time-series forecasting of COVID-19 instances for projecting effective cases using time series models, the government will be better able to deal with emergencies of this kind. Values of confirmed COVID-19 instances extremely close to the actual values may be predicted using time series forecasting data. Therefore, the primary objective of the research is to identify a model that outperforms the current models in terms of prediction and is more helpful for predicting emergencies during an epidemic or a pandemic. In the latest research, time-series models such as LSTM (Long-Short Term Memory), ARIMA (Auto Regression Integrated Moving Average), AR (Auto regression), and proposed ensemble model of Lasso regression and ridge regression with gradient boost as meta model have been studied for better Covid-19 prediction. After computing performance metrics, the root-mean-squared error, or RMSE, and mean absolute error (MAE) of each model were discovered and normalized for evaluation of the performance of the prediction models. The proposed ensemble model was found to exhibit better performance than the other models in terms of prediction accuracy. Later studies will concentrate on creating novel models capable of projecting time series data in line with the trajectory of impending COVID-19 variations.

For image-to-text and text-to-image classifications, the features of data collected from various imaging devices, sensors, and their text descriptions must be mapped into a common latent space with reduced dimensions. The low-dimensional features are supposed to provide the most information with the least amount of loss.In this paper we propose a cross-modal semantic autoencoder that uses nonnegative matrix factorization (NMF) to factorize the features into a lower rank. Due to two matrix factorization, the traditional NMF is unable to translate all of the information into lower space. This is addressed by a unique tri-factorized NMF with hypergraph regularization. Instead of using the feature adjacency matrix in hypergraph regularization, a more information-rich modularity matrix is suggested. The Wiki dataset is used to evaluate this tri-factorized hypergraph regularized multimodal autoencoder for image-to-text and text-to-image conversion. In order to lower the feature dimension, Multimodal Conditional Principal label space transformation (MCPLST) is also enabled by this novel autoencoder. Comparing the proposed autoencoder against the semantic autoencoder, the former showed an improvement in classification accuracy of up to 1.8

This research paper aims to introduce a new system that can efficiently parse video content by detecting and storing objects in each frame. The system utilizes advanced computer vision techniques to identify objects in real time, which enables users to search for specific objects based on textual queries. The system employs advanced object detection algorithms to pinpoint regions of interest in video frames and present users with an easy-to-understand visual representation. This paper provides a comprehensive overview of the system, including its design, implementation, and evaluation. The emphasis is on demonstrating the system’s effectiveness in accurately identifying and retrieving objects from extensive video datasets. The results demonstrate the system’s potential applications in video analysis, content indexing, and search-driven retrieval. The robust performance showcased in this research indicates immediate practical implications for the system, with implications for various domains. This work not only underscores the system’s utility in current applications but also lays the groundwork for further exploration and advancements in this dynamic research domain.

A collaborative real-time text editor is an application that allows multiple users to edit a document simultaneously and merge their contributions automatically.It can be made collaborative by implementing a conflict resolution algorithm either on the client side (in peerto-peer collaboration) or on the server side (when using web sockets and a central server to monitor state changes).Although web sockets are ideal for real-time text editors, using multiple collaborative editors on one connection can create problems.This is because a single web connection cannot monitor which user is collaborating on which application state, leading to unnecessary network queries and data being delivered to the wrong state.To address this issue, the current solution is to open multiple web socket connections, with one web socket per collaboration application.However, this can add significant overhead proportional to the number of apps utilized.In this study, we demonstrate an algorithm that enables using a single web socket for multiple collaborative applications in a collaborative editor.Our method involves modifying the socket's code to track which application's shared state is being worked on and by whom.This allows for the simultaneous collaboration of multiple states in real-time, with infinite users, without opening a different socket for each application.Our optimized editor showed an efficiency improvement of over 96% in access time duration.This approach can be implemented in other collaborative editors and web applications with similar architecture to improve performance and eliminate issues arising from network overload.

Technologies like self-driving cars and cleaning robots are emerging as mainstream technologies. These technologies make use of cognitive recognition. Non-negative matrix factorization (NMF) is one such technique that is popularly used for computer vision and hidden pattern recognition. NMF is prone to noises because it assumes the image signal to be linearly reconstructed. This work proposes an algorithm to increase the effectiveness of NMF and reduces the data to lower dimensions and add informational presentation which improves the clustering results of NMF. The effectiveness of the proposed model is measured by comparing them on attributes namely accuracy, homogeneity, and inertia. Some of the models that we used include K-means, PCA+K-means, NMF+K-means, Autoencoder + PCA + K-means. Our proposed model is observed to be the most effective for clustering denoised data. The algorithm also takes care of the different fault detections and gives a non-linear method based on NMF. Here, we first used autoencoders which are given input data to learn the non-linear mapping so that it can be transformed into high-dimensional space. By using the decomposition rule, we divided our feature space into two parts: The first one comprises the encoder, NMF, and decoder. This method of DNMF is a non-linear framework that can further be extended to other linear methods. The proposed method also expands the NMF's application range as it can also accept non-negative input.

Certificate management is a tedious task for any university or any other organization. These schemes impose problems in Public Key Infrastructure (PKI). Checking the validity and preserving the security of these documents is of utmost importance. In this work, we have devised a blockchain-based solution for preventing malfunctioning in certificate validation which is an important step for any university. Each certificate is uploaded in its hash format and is stored using blockchain. The hashes are stored in unique transactions in nodes, which are deployed on a private network. Using the SHA-256 hashing algorithm, the certificates are uploaded into the system and can be viewed by anyone with the right credentials. Due to the usage of blockchain technology, the certificates are stored in a decentralized manner, which ensures there is no central point of failure. Any changes in the uploaded document need to be validated by other nodes. This paper also improvises that when certificate uploading is required new nodes are added, instead of modifying the past blocks. This work provides a very user-friendly app where any user with the right credentials can upload documents. In this work, digitized documents are stored using Inter Planetary File System (IPFS) which is distributed method of storage. Our theoretical analysis proves that it is a user-friendly application with the security of blockchain technology in partnership with IPFS. Only the issuer can upload documents and others can only view them. Using our proposed solution, problem of malicious certificates can be tackled with E-certification. The proposed method solves all the issues of storing, validating, and sharing documents. Chaotic Map technique is used in hash generation which is quite simple to implement. The proposed approach Chaotic Key based Certificate validation (CK-Cert) provides a hassle-free solution for certificate managements since it better manages the block size as compared to previously proposed techniques (PBCert and CertChain) as discussed with the help of graphs.

Regression testing of the software during its maintenance phase, requires test case prioritization and selection due to the dearth of the allotted time. The resources and the time in this phase are very limited, thus testers tend to use regression testing methods such as test case prioritization and selection. The current study evaluates the effectiveness of testing with two major goals: (1) Least running time and (2) Maximum fault coverage possible. Ant Colony Optimization (ACO) is a well-known soft computing technique that draws its inspiration from nature and has been widely researched, implemented, analyzed, and validated for regression test prioritization and selection. Many versions of ACO approaches have been prolifically applied to find solutions to many non-polynomial time-solvable problems. Hence, an attempt has been made to enhance the performance of the existing ACO_TCSP algorithm without affecting its time complexity. There have been efforts to enhance the exploration space of various paths in each iteration and with elite exploitation, reducing the total number of iterations required to converge to an optimal path. Counterbalancing enhanced exploration with intelligent exploitation implies that the run time is not adversely affected, the same has also been empirically validated. The enhanced algorithm has been compared with the existing ACO algorithm and with the traditional approaches. The approach has also been validated on four benchmark programs to empirically evaluate the proposed Enhanced ACO_TCSP algorithm. The experiment revealed the increased cost-effectiveness and correctness of the algorithm. The same has also been validated using the statistical test (independent t-test). The results obtained by evaluating the proposed approach against other reference techniques using Average Percentage of Faults Detected (APFD) metrics indicate a near-optimal solution. The multiple objectives of the highest fault coverage and least running time were fruitfully attained using the Enhanced ACO_TCSP approach without compromising the complexity of the algorithm.

Abstract The data from different sensors, cameras, and their text descriptions needs their features to be mapped into a common latent space with lower dimensions for image-to-text and text-to-image classifications. These low-dimensional features should incur maximum information with minimum losses. The cross-modal semantic autoencoder is proposed in this paper, which factorizes the features into a lower rank by nonnegative matrix factorization (NMF). The conventional NMF lacks to map the complete information into lower space due to two matrix factorization which is overcome by a novel tri-factor NMF with hypergraph regularization. A more information-rich modularity matrix is proposed in hypergraph regularization in place of the feature adjacency matrix. This tri-factorized hypergraph regularized multimodal autoencoder is tested on the Wiki dataset for the image-to-text and text-to-image conversion. This novel autoencoder is also supported by Multimodal Conditional Principal label space transformation (MCPLST) to reduce the dimension of the features. The proposed autoencoder observed a classification accuracy improvement of up to 1.8 % than the semantic autoencoder.

Objectives: Software researchers have been taking advantage of various evolutionary optimization approaches by digitizing them. Test case selection and prioritization based on fault coverage criteria within a time-constrained environment is important in regression testing problem. Methods: This work empirically evaluates different approaches that includes evolutionary approaches (Ant Colony Optimization, Bee Colony Optimization, a combination of Genetic Algorithms and Bee Colony optimization), and a Greedy approach. These tetrad techniques have been successfully applied to regression testing. Also, tools have been developed for their implementation. Eight open-source test programs, written in C language have been used for empirical evaluation of the regression testing approaches. Findings: The accuracy achieved by t-GSC, being a greedy technique, was found to be least; while that of ACO was found to be the best. All the tetrad approaches yielded borderline better or worse results, while all the four gave excellent time and size gains. Novelty: There are many studies available in the literature that compare various regression testing approaches of a similar kind. Instead of repeating the same, it is intended to evaluate two well-accepted approximation approaches: a hybrid approach, and a greedy approach. It has been tried to evaluate the efficiency of the greedy approach with the metaheuristic approach. It is imperative to compare approaches following different algorithmic paradigms, yet trying to solve the same problem. Keywords: Ant Colony Optimization; Bee Colony Optimization; Genetic Algorithms; Greedy Set Cover; Software Testing; empirical comparison

Cross-modal retrieval intends to empower adaptable recovery across various modalities. The center of cross-modal retrieval is the manner by which to quantify the substance similitude between various sorts of information. In this work, we deal with a cross-modal retrieval technique, called Canonical Correlation Analysis (CCA). It accepts one sort of information as the question to recover pertinent information of another sort. The given indexed lists across different modalities can be useful to the clients to get exhaustive data about the objective occasions or points. With the quick development of various kinds of media information like texts, pictures, and recordings on the Internet, cross-modal retrieval turns out to be progressively significant in true applications. As of late, cross-modal retrieval has drawn in the significant consideration of the analysts from both scholarly communities also, industry. The test of cross-modal retrieval is the ticket to gauge the substance closeness between various kinds of information since they, which is alluded to as the heterogeneity hole. After data preprocessing and learning the mappings in the same space, we will try to find out the most similar samples based on pre calculated features of the samples in a given format. We will keep the features learned by VGG-16 precalculated and the features learned by text model would then be used to search for the most similar image that best explains the caption.

The present investigative study is a systematic as well as scientific endeavor to understand the beneficial effect of the text mining approach for the service management of the organization. This stud uses secondary qualitative data for gaining accurate insight into the present research study. Thematic analysis is done on the collected data set for gaining accurate insight into the present study. Interpretive research philosophy is chosen for this study along with the inductive research approach. This study also uses an explanatory research design for effective insight into the beneficial effect of the text mining process for the service management of the organization. The study focuses on the process of text mining which is used for analysing the client forums, call logs, tickets of customer service, and platforms of social media, news feeds, tweets and many others. It also focuses on the business that have better insights based on client expectations and it requires the areas that focus on overall improvements. The study integrates the customer communications and reviews that can identify the new features for strengthening the product offerings. It focuses on the latest technologies that provide several opportunities for improving the customer experience those results in increased profits and revenue.

The real-world data is multimodal and to classify them by machine learning algorithms, features of both modalities must be transformed into common latent space. The high dimensional common space transformation of features lose their locality information and susceptible to noise. This research article has dealt with this issue of a semantic autoencoder and presents a novel algorithm with distinct mapped features with locality preservation into a commonly hidden space. We call it discriminative regularized semantic autoencoder (DRSAE). It maintains the low dimensional features in the manifold to manage the inter and intra-modality of the data. The data has multi labels, and these are transformed into an aware feature space. Conditional Principal label space transformation (CPLST) is used for it. With the two-fold proposed algorithm, we achieve a significant improvement in text retrieval form image query and image retrieval from the text query.

Composing music is a very interesting challenge that tests the composer’s creative capacity, whether it a human or a computer. Although there have been many arguments on the matter, almost all of music is some regurgitation or alteration of a sonic idea created before. Thus, with enough data and the correct algorithm, deep learning should be able to make music that would sound human. This report outlines various approaches to music composition through Neural Network model, it is evident that musical ideas can be gleaned from these algorithms in hopes of making a new piece of music. The use of deep learning to solve problems in literary arts has been a recent trend that has gained a lot of attention and automated generation of music has been an active area. This project deals with the generation of music using some form of music notation relying on various LSTM (Long Short Term Memory) architectures. Fully connected and convolutional layers are used along with LSTM's to capture rich features in the frequency domain and increase the quality of music generated. The work is focused on unconstrained music generation and uses no information about musical structure such as notes or chords to aid learning.

The growth of the internet and the social media has led to rapid growth of the digital images available online as digital databases. The problem of searching such databases for items that are similar to a query image or a query descriptor is termed as image retrieval, and such systems are called image retrieval systems. An image retrieval system is a computer system that helps in looking up a large database of digital images and retrieving the images, which matches a user’s request. These systems are used for retrieving images related to the user request from the database. The following research will state the role of generating descriptions for images in these systems and proposing the use a multilayer biLSTM (bidirectional LSTM) for the purpose of generating descriptions. The proposed model gave desirable results, which were also better results than its counterparts, i.e., LSTM and biLSTM.

In this paper, we introduce a framework that simplifies the process of finding information about a city. This is helpful when people are looking for information on various cities before planning a trip. As visiting various online portals and collecting information is a time consuming and tedious task , we design a service that automates this task. We also provide reviews and analyze them using sentiment analysis. This will provide better insights to the user. The text database is generated dynamically using web scraping hence the quality of final results depends on the quality of content available online.

The primary aim of cross-modal retrieval is to enable the user to retrieve data across different modalities in a flexible manner. Through this paper, we tackle the problem of retrieving data across different modalities, where the input is given in one form, and relevant data of another type is retrieved as the output, as per the requirement of the user. Most of the techniques or approaches that have been used so far have not considered the feature and semantic information preservation. As a result of this negligence, they are not able to obtain effective results. Here, we have proposed a two-stage learning method that does the projection of low dimensional embeddings to multimodal data that preserve both feature and semantic information, which enabled us to get satisfactory results. In this paper, we have proposed an autoencoder for cross-model retrieval that can process both visual as well as textual data based on their semantic similarity.

Background: The medical data, in the form of prescriptions and test reports, is very extensive which needs a comprehensive analysis. Objective: The gene expression data set is formulated using a very large number of genes associated to thousands of samples. Identifying the relevant biological information from these complex associations is a difficult task. Methods: For this purpose, a variety of classification algorithms are available which can be used to automatically detect the desired information. K-Nearest Neighbour Algorithm, Latent Dirichlet Allocation, Gaussian Naïve Bayes and support Vector Classifier are some of the well known algorithms used for the classification task. Nonnegative Matrix Factorization is a technique which has gained a lot of popularity because of its nonnegativity constraints. This technique can be used for better interpretability of data. Results: In this paper, we applied NMF as a pre-processing step for better results. We also evaluated the given classifiers on the basis of four criteria: accuracy, precision, specificity and Recall. Conclusion: The experimental results shows that these classifiers give better performance when NMF is applied at pre-processing of data before giving it to the said classifiers. Gaussian Naïve Bias algorithm showed a significant improvement in classification after the application of NMF at preprocessing.

With increasing number of users contributing in digital text generation, it becomes necessary to have some tools to extract relevant information from these text documents. Manually performing this task is a time consuming and tidious process. Both Matrix Factorization Based and Graph-Based Models are unsupervised models for extracting a summary from text documents. As the training phase is not required for both these methods, they are extremely fast. Both these methods construct an intermediate representation of a text document and use it to assign a score to each sentence present in the text document. In this paper, we discussed the underlying concept behind both the methods and compare them on the basis of quality of summary extracted.