Enhancing the brain involvement, attention, and memory of students has been proven to be very feasible with the introduction of music for classroom learning environments. This research proposes for Intelligent Music Recommendation System (IMRS) which can be used to change the background music according to the state of the brain, tasks and tastes of trainee. A mixed method is used for the study, which consists of a combination of quantitative analysis of learning success measures and personal reviews of student experience. Machine learning and deep learning algorithms, Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM), K-Nearest Neighbours (KNN) and Support Vector Machines (SVM) are adopted in the proposed system to look at the music features and guess the best soundscapes for different learning situations. Information about each student, such as age, subject, attention level and mood is combined with information about the environment to make personalised music suggestions. To ensure that they meet the needs for cognitive load, music files are processed for the purpose of emotional mapping, feature extraction and tag extraction. When compared to traditional background music methods that don't change, the IMRS framework is judged on how well it helps people to focus, relax and remember things. The consequences of the predicted results for the educational experience demonstrate the usefulness of AI-driven personalisation in educational contexts by showing how smart music selection can help students do better in school and be happier with their lives. This is a new combination of artificial intelligence, music psychology and educational technology, which has made learning more fun and more effective in the modern day world.

In the modern digital era, textual data plays a pivotal role in the dissemination of information through various forms such as news articles, reports, and papers. With the exponential growth in the production of such content, there arises an urgent need to efficiently extract meaningful and insightful knowledge from this vast pool of text. natural language processing (NLP) has been developed to tackle this challenge, enabling machines to comprehend textual data and derive valuable information. Text summarization, a critical NLP task, involves condensing large volumes of text into shorter, more digestible summaries while retaining key factual data. This process aids in managing the overwhelming influx of information and makes it more accessible and actionable. In this paper, the research aims to implement a text summarization system using the BART (Bidirectional and Auto-Regressive Transformers) model, augmented with various ranking-based algorithms, to enhance the quality and relevance of the summaries. This approach seeks to introduce a human-like touch to the summarization process, ensuring that the generated summaries are not only concise but also contextually accurate and meaningful for the end-user.

As cyber threats become more complex, real time systems are needed to detect and eliminate attacks. Traditional network intrusion detection systems based on rule based static method tend to be ineffective against novel emerging threats. In this paper we propose an improved real time cyber threat detection system using adaptive machine learning techniques used to analyze network traffic and find anomalies. Our proposed approach uses a blend of supervised and unsupervised learning models such that the system maintains high detection accuracy with minimal false positives, while maintaining continuous adaptation to constantly evolving threats. On critical network traffic features like packet size, flow duration, source and destination IP addresses, transmission protocols, the system is then trained. They show experimentally better detection accuracy, responsiveness and adaptability than conventional IDS. In this work, contributions of adaptive machine learning for robustness against dynamic and evolving threats in network environments are highlighted as significant strides towards improving real time cybersecurity infrastructure.

In the realm of fashion technology, the introduction of smart textiles and wearable sensors may be attributed to the fast progression of technology and the rising need for customized, interactive, and intelligent apparel. This study investigates the interface of electronics and textiles to develop cutting-edge, multifunctional, and flexible apparel. This research digs into the fascinating world of smart textiles. It investigates the many different kinds of sensors, such as temperature sensors, heart rate monitors, motion detectors, and chemical sensors that may be easily incorporated into the fabric. These sensors make a wide variety of applications possible, such as improved user experiences, monitoring of health conditions, and data collecting on the surrounding environment. In addition, to assure the practicability and usage of smart textiles, it addresses the problems and possibilities connected with various power sources, connection choices, and data processing units. The purpose of this study is to contribute to the continuous development of intelligent clothing, which has the potential to revolutionize the fashion industry by producing clothes that not only represent an individual's sense of style but also give useful insights, functionality, and connectedness to the digital world.

Abnormal behavior methods have attempted to reduce execution time, computational complexity, efficiency, robustness against pixel occlusion, and generalizability. This research proposed a novel method in human activity-based anomaly detection and recognition by surveillance video utilizing DL methods. Input is collected as video and processed for noise removal and smoothening. Then kernel local component analysis extracts these video features for human activity monitoring. Then the extracted features are classified using Bayesian network-based spatiotemporal neural networks. The classified output shows the anomaly activities of the selected input surveillance video dataset. The simulation results are obtained for various crowd datasets regarding the mean average error, mean square error, training accuracy, validation accuracy, specificity, and F_measure. The proposed technique attained an MAE of 58%, MSE of 63%, specificity of 89%, and F-measure of 68%. training and validation accuracy of 92% and 96% respectively.

One of the helpful tool for measuring and managing home power consumption is one of the clever and creative tool for tracking the usage of Home Power Management System. The Goal of this system is to lower the power bills, also promote energy efficiency, and provide the users active control over how much energy they use. Smart meters, an intuitive and user-friendly smartphone app, and real-time energy monitoring sensors are some of the system's essential parts and these are the key parts of the system. Installing smart meters makes it possible to accurately track the power usage and provide data to the system. Users can be able to see their energy usage instantly and make smart decisions when real-time energy monitoring equipment is included. Further, the system uses advanced smart technology, analytics and machine learning algorithms to provide clients customized suggestions on how to save energy consumption. Homes may save a lot of money on electricity bills, help the environment, and have more control over how much energy they use by using our Smart Home Power Management System.

Music genre classification, the task of automatically assigning audio recordings to specific genres, offers exciting possibilities for music recommendation systems, personalized playlists, and music analysis. This abstract explores the potential of (K-NN), a versatile Artificial Intelligence (AI) algorithm, in tackling this challenge- NN excels in its simplicity and interpretability. It classifies new music by identifying its closest neighbors within a labeled dataset based on predefined features, such as tempo, timbre, or spectral content. These neighbors then collectively "vote" on the genre of the new piece. The abstract delves into the feature engineering process, highlighting the crucial role of selecting relevant musical characteristics for accurate classification. Furthermore, the abstract discusses the impact of the "K" hyper parameter, which determines the number of neighbors considered in the voting process. Choosing the optimal "K" value balances accuracy and sensitivity to genre nuances. While K-NN offers advantages in ease of implementation and understanding, the abstract acknowledges its potential drawbacks. The storage requirements can be high for large datasets, and computational costs may increase with data size. It concludes by acknowledging the need for further research to address computational challenges and explore advanced feature engineering techniques to unlock the full potential of K-NN in this exciting domain.

Effectively forecasting and reducing the effects of sudden weather- related catastrophes require sophisticated weather monitoring and disaster mitigation systems. This essay provides an overview of such a system. The proposed system collects real-time weather data and reliably forecasts the occurrence of severe weather events by utilizing cuttingedge technologies like remote sensing, satellite images, and machine learning algorithms. The system can detect and foresee the creation and movement of weather patterns, such as hurricanes, tornadoes, and heavy rainfalls, by continually monitoring atmospheric parameters, including temperature, humidity, wind speed, and air pressure. Early detection permits prompt action and the right steps, such giving warnings, ordering evacuations, or launching disaster assistance operations. In order to evaluate the probable impact of a disaster event, the system also contains a thorough disaster mitigation approach by combining several data sources, such as topographical maps, population density statistics, and infrastructure vulnerability evaluations. By identifying high-risk regions, devising targeted mitigation plans, and putting preventative measures in place to reduce damage and casualties, this information supports the development of mitigation strategies. Ultimately, the goal of the advanced weather monitoring and mitigation system is to protect people, and economies by improving preparedness, response, and recovery activities.

Using text mining tools and machine learning algorithms, the paper presents a prototype for classifying strokes. The significance of machine learning extends across various domains, including surveillance, medicine, and data management, where appropriately trained algorithms prove invaluable. The study incorporates data mining techniques, offering a comprehensive understanding of information tracking from both semantic and syntactic perspectives. By analyzing medical case sheets, the proposed approach extracts symptoms and trains the system based on that information. A total of 507 case sheets are collected at "Sugam Multispecialty Hospital" in "Kumbakonam", Tamil Nadu, India, during the data collection phase. A tagging and maximum entropy approach was used to mine these case sheets. Stroke classification relies heavily on the suggested stemmer for distinguishing between common and special features. To handle the data that has been processed, support vector machines, XG Boost algorithms, SGD algorithms, decision trees, and random forests were used as machine learning algorithms. Random Forest demonstrated the highest accuracy among stroke classification algorithms evaluated. This integration of text mining and machine learning offers a promising framework for enhancing stroke classification based on patients' symptoms extracted from medical case sheets. The synergy between these technologies provides a robust methodology that can contribute significantly to improving the efficiency and accuracy of stroke diagnosis and classification in a medical.

Nowadays, digital withdrawals mostly rely on cards. Days have come where people don't have to carry cash in their pockets and just a small card is enough to make all the withdrawals. Problems with cards will lead to passwords being hacked and easily lead to fraud. Credit card fraud can be identified using Hidden Markov Model (HMM) and Formula Based Authentication. In the Existing system, card withdrawals are very routine among the people and the frauds corresponding to the improvement of security are increasing. In the proposed system, ML algorithms have been applied to detect MasterCard deception in a disproportionate dataset. In the modification process, an application is developed for a banking sector particularly for a credit or ATM card. Users can create an account and get the ATM card along with a unique formula which should be used during suspicious transactions. The user behavior of every transaction is tracked by Hidden Markov Model and if there are any occurrences of suspicious transactions, then a message is sent to the user with the keys that are required to complete the formula. After the user applies the keys to the formula the solution must be entered as the password in order to complete the transaction.

The human race has developed various automobiles for faster transportation from one place to another place. But these automobiles are the main cause of human loss in the current trend. Various reasons cause these accidents, out of which three main reasons are accidental door openings, unforeseen lane shifts, and blind spot misses during parking. These accidents are persistent and happen very often, posing a severe threat to drivers, pedestrians, other drivers, and pavilion passengers, if luck is on our side, it would just pause at the cost of vehicle damage. To prevent all these occurrences, the “Integrated Door Opening And Lane Change Warning System” was introduced. This Integrated Door Opening And Lane Change Warning System is an apparatus that keeps a constant eye on the surroundings at all times. It is an integration of sensors that is installed into the vehicle which checks for any approaching vehicles, pedestrians, or bikers that are in the circumference of it. If the apparatus notes the presence, it takes action accordingly. Rather than acting upon the incident after its occurrence, this system tries to prevent it, upon detecting the contingency of an accident. Even if the passenger catches a blind spot, the Integrated Door Opening And Lane Change Warning System has your back, keeping track of all the surroundings and letting everyone in the vehicle at ease.

Considering the challenges inherent in the removal of waste particles from water surfaces, it is imperative to address this issue with a comprehensive and strategic approach to ensure effective and environmentally sustainable water cleaning solutions.This paper suggests designing a system that employs a boat convoyed belt ocean floor garbage removal to clear ocean garbage easily along with a tracking and a monitoring unit to monitor the weather. An Arduino Uno microcontroller with the ATmega328p chip is used to collect and manipulate sensor data.TheDHT11 sensor, flame sensor, ultrasonic sensor, water moisture sensors are utilizedby microcontroller. GPSsensor is used to locate the boat and thelocation is collected and uploaded to the IoT cloud. LoRa uses extremely little power and is utilized for IoT device connectivity.The formatted sensor and GPS location data are transmitted using the LoRa module. A LoRa gateway configuration for the LoRa module is required.The system's microprocessor instructs the buzzer to sound an alert in situation that requires attention. Blynk app is used to check the insights of the sensors and to control boat using Bluetooth

The conversion process of words to vectors involves mapping each word or term in the biomedical data to a unique numerical vector, called a word embedding. One of the key advantages of using word embeddings is that they capture the meaning and context of words in a numerical format, which can be used as input for various machine learning and data analysis tasks. The Milvus vector engine is an open source software that allows for the efficient and accurate representation of large-scale data in vector space. It is particularly useful for high-dimensional data such as biomedical data that contains a large number of unique words and terms. The use of the Milvus vector engine is to convert biomedical data into numerical vectors for machine learning and data analysis tasks. In the context of biomedical data, the use of word embeddings can enable new discoveries and insights by allowing for more accurate and efficient analysis of large datasets. This word embeddings can be used to identify patterns and relationships in biomedical literature, or to classify medical articles based on their content.

Makaton is a widely used sign language system that is primarily used by individuals with communication difficulties, such as those with Autism, Down syndrome, and other developmental disorders. However, despite its widespread usage, there is a lack of technology for accurately recognizing and understanding Makaton signs. This paper proposes a model that combines MediaPipe Holistic with a neural network, such as SimpleRNN, LSTM, or GRU, to recognize Makaton sign language (MSL). The model is trained on a custom dataset of Makaton sign language videos using backpropagation and the Adam optimizer. The proposed network achieved a high accuracy on the test set, demonstrating its effectiveness in recognizing Makaton sign language. This study introduces an innovative method for advancing the development of systems that recognize sign language by utilizing MediaPipe Holistic in conjunction with a neural network. This can improve the way technology interacts with people who have hearing impairments.

Convolutional Neural Networks (CNN) are used widely adopted for tasks involved with Computer Vision, Medical Imaging and Natural language processing. Creating a CNN model which has the ability to detect and track objects as similar to human remains a challenging task. CNN can perform well if the object to be detected is found in the same position as the one found in the dataset. But, if the object contains certain angle of tilt or rotation, then detecting the object remains a tedious or laborious task. The task of image classification has 2 stages namely Feature extraction and classification. In this paper, we have deployed 2 techniques data augmentation and Transfer learning and the comparisons are studied. Data Augmentation is a technique adopted for rearranging the images in different orientation by applying scaling, rotation and shearing. Through Transfer learning pertained model is applied for feature extraction which can then be applied over the new model. The dataset considered for our approach is CIFAR10 dataset. Applied with CNN, we have obtained an accuracy of 84.56%. The model after applying Data Augmentation has received 96.56% accuracy in which 20% of data is considered for drop out. The model is trained with Transfer learning RESNET 50 architecture and the validation process has received an accuracy of 98.79. The prediction accuracy with Data augmentation process is 96.56% and with transfer learning is 100% as it is well pre trained model applied over the new dataset.

Lithiumion batteries are crucial to the functioning of contemporary society, since they are used in a wide variety of applications from powering electric cars to storing renewable energy. Yet, there is still a significant problem in making these batteries safe and long-lasting. Using an unique CNN-LSTM-DNN architecture, we offer a new approach for predicting the RUL of lithium-ion batteries in this research. Our method combines the strengths of CNNs for image analysis, LSTMs for sequence data, and DNNs for feature learning. By combining these models, we are able to outperform existing state-of-the-art approaches to estimating battery RUL. An extensive lithium-ion battery dataset from the Center for Innovative Product Lifecycle Management is used to assess the efficacy of our approach; the findings show a considerable decrease in identification mistakes and an increase in RUL identification performance. This research provides the path for a more accurate and efficient early warning system to be created, which will be crucial in preventing dangerous and short-lived lithium-ion battery failures in the future.

Enterprise Resource Planning (ERP) are programmed for institutionalizing, streamlining and coordinating business forms crosswise over various account, acquisitions, for circulation purpose, and for other different divisions. The day e-organizations have begun to rise, organizations are following the ways to be progressively profitable. The objective of this paper is to optimize the critical factors such as cost and time in various ERPs. In this paper, the Enhanced Recursive Feature Elimination method has been proposed for feature selection. Modeling of ERP parameters such as Design of Experiments (DoE) and Response Surface Methodology (RSM) are done with linear regression and the significance of the model is studied using ANOVA table. Feature selection is done also done using Genetic algorithm. Comparisons are made between Manual planning and Genetic algorithm based Recursive Feature selection method and experimental results have shown ERP cost can be reduced to 40% by adopting the proposed approach.

With the rapid growth of the technology, nowadays more IoT devices are enabled and more data is transferring from one device to another device. Due to huge data transmission from one node to another a traffic or collision may occur in IoT Assisted Smart devices in WSN. To avoid the issues an energy efficient routing protocol is used. In this regard, routing is a challenging task while transmitting the packets to some storage repositories. This paper presents the Energy Efficient Routing Approach (EERA) to identify the routing paths where less energy is consumed to enhances the network lifespan.

Missing data imputation is an ongoing and crucial research topic in data mining. There may be many missing values in large dataset. However, there are few methods used solely for downstream analyses, with a few prediction tools, which definitely do need a full descriptor value matrix. We propose and assess a looping imputation method called ensemble based on few imputation methods. By calculating an average over a lot of regression trees which are unpruned, the ensemble method intrinsically constitutes a multiple imputation scheme. Using bagging estimate, boosting estimate and stacking estimate of the ensemble method, we are able to estimate the imputation error. Evaluation is finished on molecular descriptor datasets generated from a diverse choice of pharmaceutical fields with artificially delivered missing values ranging from 10 to 30%. The experimental end result demonstrate that missing values have an amazing impact at the effectiveness of imputation strategies and our approach ensemble is sturdier to missing values than the alternative ten imputation strategies used as benchmark. Additionally, the ensemble method exhibits appealing computational performance and can address high-dimensional data.