A key component of software quality assurance is software error prediction, which aims to find probable flaws in software systems during the development stage. In this extensive study, we assessed the classification efficiency of four well-known classifiers for software defect prediction: Decision Tree, Multi-Layer Perceptron, Random Forest, and Multi SVM. Additionally, we revised the firefly method and developed a unique multi-objective fitness function that took into account both weighted error and area under the curve (AUC) in order to assess the fitness of the classifiers in feature selection. Using a diverse dataset comprising varying numbers of records, ranging from 500 to 5000, we calculated key performance metrics, including precision, recall, F-measure, classification accuracy, and receiver operating characteristic (ROC) curves. The average metrics for each classifier at 5000 records were as follows: In terms of precision, recall, F-measure, and accuracy, DT attained averages of 0.7974, 0.8133, and 0.8047. Similarly, MLP exhibited an average precision, recall, F-measure, and accuracy of 0.8130, 0.8041, 0.7984, and 0.7953, Random Forest exhibited 0.8138, 0.8047, 0.8172, and 0.7950, while Multi SVM exhibited 0.8032, 0.8117, and 0.8128, respectively.

Background:Spams are commonly known as unwanted commercial or deceptive emails, which strategically target specific individuals or businesses to promote products or mislead recipients. However, with the implementation of advanced technologies such as machine learning and natural language processing, computers can be trained to discern and categorize these emails as spam or legitimate (ham) messages. Despite considerable efforts in spam filtering, the effective identification and mitigation of spam emails remain an ongoing challenge. Methods:This research places particular emphasis on scrutinizing email headers and extracting crucial data, such as HOP count and IP address, using a Python script that serves as a forensic or investigative tool for analyzing and extracting information from email files. Additionally, it assesses various vectorization techniques to gauge the efficacy of machine-learning approaches for spam classification. The work encompasses a range of supervised learning algorithms, including Logistic Regression, Decision Trees, Naive Bayes, and Natural Language Processing (NLP) methods, such as Bidirectional Encoder representation of transformers (BERT). Two vectorization methods, count vectorization and tf-idf vectorization, are compared. The evaluation metrics employed included accuracy, training time, CPU and wall times, precision, recall, f1 score, and support. Conclusion:The performance of the Decision Trees is particularly noteworthy, achieving a flawless 100% accuracy rate. The trained model is seamlessly integrated into both an Android application and a website, enabling real-time spam detection and classification.

Encoder-decoder-based video captioning gives a holistic description as per the training data. These captions are missing the object’s motion-specific features. Object motion knowledge in video allows object-oriented video captions. Similarity action knowledge-based models allow action-based features. Furthermore, the traditional encoder-decoder method uses frame-level scene features. Advanced existing methods extract spatial and temporal features for extracting context vectors. The unavailability of methods to extract action knowledge with the object’s motion features limits the models to produce action-object-oriented captions. Further presence of multiple objects’ motion gives disoriented captions in state-of-the-art methods. We propose a method that is a partial grid-based method for action-object-oriented features. This facilitates comprehension of an object's motion and its interactions with other objects, as well as movement within the scene. The proposed method takes these features and constructs a graph neural network, which is then used with graph-based filters. Object activity and interaction based re-annotated 75 videos from MSVD datasets which were used for training, validation, and evaluation. The proposed model demonstrates object-action-based video captioning with object-action and object-background interaction. The BLEU and METEOR-based evaluation results demonstrate the workability of graph neural network-based methods and the superiority of the process.

Nowadays, social media platforms are extensively used by the public to expose their opinions on various sensitive matters. One of the active research challenges in social media analytics is web content mining and context analysis. This helps us to identify events or incidents which are actively discussed among many users who may be from specific geographical areas. Such events or incident identification gives an early warning in many unusual situations [1]. However, the semantic processing of social media is challenging due to its high complexity, ambiguity, and unstructured nature. In this work, we propose a framework for crowd cluster identification and context analysis from clusters using the Online Spherical K-means algorithm and some Natural Language Processing (NLP) techniques. Initially, the tweets are scraped from Twitter and undergo suitable data preprocessing steps. Furthermore, clusters are identified from the cleaned data using the Online Spherical K-means algorithm. Finally, the analysis and visualization of context discussion from each cluster are performed with the aid of various fitting NLP methods. The proposed method is evaluated using tweets scraped with three different hashtags #blacklivesmatter, #Superbowl, and #Texasfreeze. For performance evaluation, we computed the homogeneity score, Completeness score, Calinski-Harabasz Index, and V-Measure. The performance metrics show that the proposed method yields promising results.

Within the realm of sustainable agriculture and the protection of crops, the presence of intruders poses a substantial risk, leading to potential crop damages and injuries. In our paper, we conceptualize intruders as objects. Consequently, the implementation of object detection emerges as a crucial safety and field protection measure, effectively mitigating the adverse consequences of such encounters. While deep learning techniques have proven to yield superior results in object detection, their computational and parameter requirements have posed challenges in their widespread implementation. This research paper presents an alert system AgriGuard equipped with a lightweight object detection model for enhancing security in agriculture fields. The proposed embedded system utilizes EmbdYOLOv3 and TinyEmbdYOLOv3 models, modified versions of YOLOv3 and TinyYOLOv3, respectively, to overcome challenges in object detection, such as occluded or look-alike animals. The system integrates ultrasonic sensors, raspberry pi, real-time object detectors, google firebase, and an android application to detect and alert farmers for unauthorized access. Experimental results show that EmbdYOLOv3 outperforms YOLOv3 by 36.85%, and TinyEmbdYOLOv3 outperforms Tiny-YOLOv3 by 14.48% under the same embedded environment. The study highlights the effectiveness of low-power IoT devices and deep learning techniques in providing robust security solutions for the agriculture industry, offering a potential approach for mitigating crop damages and ensuring field safety.

The industry 4.0 paradigm introduces information and communication technologies (ICT) that disperse information processing and decentralize decision-making to several autonomous and intelligent entities, such as production resources, operators, and products. To address Industry 4.0's decentralized decision-making and complex production systems, we developed a multi-agent simulation model that enables risk-free experimentation and performance optimization. This article presents a multi-agent system model; the proposed model is a decision support tool that allows simulation, evaluation, and amelioration of the performance of the production system. This decision support system aims to analyze logistics strategies, calculate production rates, and simulate different scenarios. Our results demonstrate its effectiveness in evaluating and improving production system equipment reliability, availability, and performance.

Recent developments in deep learning have revolutionized the field of activity recognition by humans and other entities. These advancements allow models to learn complex representations and hierarchies from raw data, hence increasing recognition accuracy. Many machine learning algorithms, such as support vector machines and histogram of gradients with k-nearest neighbor classifiers, have lost part of their appeal due to the extensive feature engineering and data preparation required. The objective of this research is to construct an efficient and robust ensemble model by utilizing the strengths of Convolutional neural networks (CNN), Visual Geometry Group (VGG16), Inception model, and Residual Networks 50 (ResNet50) model in order to boost the resilience and predicted accuracy of human activity recognition from raw visual data. This work makes use of a large and diverse dataset that includes over 12,000 annotated photos illustrating various human activities and methodology yields promising results while removing the need for sophisticated feature manipulation. Furthermore, the results show how the ensemble model is better at Human Action Recognition (HAR).