Background: LLMs are quite good, but the problem is that they are constrained by the data they use. This means that they might give false information or create untrue statements. This has led to a new technique called Retrieval-Augmented Generation (RAG), which aims at addressing this problem by providing LLMs access to information outside their dataset. Most of these techniques are only able to use text data, but this text might not be current.Methods: We created DataMind AI to solve a particular problem. Our system works as an intelligent assistant that can analyze content, both text and web pages, to retrieve information. We use special search techniques that ensure a good balance of thoroughness and speed. Moreover, a part of our system, called the ’Online Retrieval Booster,’ helps retrieve information from the internet when needed, thus improving the overall accuracy of our system. Our Retrieval-Augmented Generation (RAG) model uses a tree-based data structure to store contextual similarity, thus reducing the need to parse vectors in the database.Results: The evaluation was done to test the ability of DataMind AI to answer challenging queries. From the results obtained, it is clear that DataMind AI offers better responses compared to the usual AI and other similar technologies. In addition, the technology has fewer errors since it checks the information from reliable and updated sources, hence providing more precise and useful responses that are in line with the question.Conclusion: The DataMind AI addresses the problem of the limited knowledge of language models in a robust and scalable way by providing a solution for the effective combination of local multimodal data and web information, thus creating a path toward more reliable, auditable, and intelligent AI systems.

Artificial intelligence (AI) integration in programming education can improve learning efficiency, provide individualized feedback, and offer adaptive assessments for students of different abilities. Most current coding platforms focus on competition among students instead of offering intelligent tutoring, graphics visualization tools, and collaborative capabilities. This paper compares three supervised machine learning algorithms to predict student programming success and future AI-assisted learning in the classroom: Support Vector Machines (SVM), Random Forests (RF), and Artificial Neural Networks (ANN). Using datasets collected from the UCI Student Programming Performance repository and IBM CodeNet, we trained and evaluated each model based on the programming characteristics that are lexical, structural, and behavioral. The ANN model performed best, achieving 93.60% accuracy, 0.94 precision, 0.93 recall, and 0.93 F1-score—making it superior to both RF and SVM. Our research suggests that ANN models excel in identifying nonlinear and complex patterns of learning, making them the optimal choice for intelligent adaptive coding educational platforms. In the future, we will build an integrated AI-driven collaborative coding platform that includes personalized tutoring and real-time feedback.

The rapid digitization of healthcare has significantly enhanced the access to patient data, but has also raised significant concerns related to data privacy, ownership, and secure sharing. Typical Electronic Medical Record systems generally store patient’s data on servers managed by a hospital and lacks full transparency and independence for patients. This lack of transparency often leaves patients with no understanding of who has accessed to their record, along with the inability to efficiently search their unstructured medical data. To address these challenges, this paper proposes a Secure Patient-Driven Medical Record Sharing and Search System to restore full control of medical data back to patients. This framework allows the patient to encrypt, manage and share his/her health records with information encoded in Advanced Encryption Standard encryption and hashed metadata to safely search while preserving privacy. Access logs maintain transparency, along with an associated Artificial Intelligence-based doctor recommendation module that processes patient symptoms and suggests specialists without revealing their identity. The experimental results indicate that the system provides enhanced data control, usability, and search efficiency over current Electronic Medical Records systems.

Aspect Based Sentiment Analysis (ABSA) aims to determine sentiment with respect to specific aspects of a text, providing more detailed insights than conventional sentiment analysis which assigns a single polarity to the whole text. However conventional techniques often fail to capture fine grained aspect level sentiment limiting their effectiveness in real world applications such as product reviews and customer feedback analysis. In this article proposes an innovative ABSA framework that synchronizes enhanced feature engineering with a lightweight hybrid deep learning architecture. In the proposed method Text preprocessing is done using a BERT tokenizer followed by feature extraction with an improved TF-IDF approach and Aspect Term Extraction (ATE) allowing the model to capture both global context and aspect level information. A hybrid classifier synchronizes Link-Net and SqueezeNet for fast and accurate sentiment classification. Experiment was conducted on data for Restaurant Reviews containing 10,000 reviews. The dataset splitting in training, validation and testing set 70:15:15 respectively. Fiive-fold cross-validation was applied to ensure robustness of proposed framework. Outcome of analysis demonstrate that the proposed farmwork gained a high accuracy of 97.7

Electronic devices and internet purchasing are more common today. For online line shopping most of people are using internet banking and credit for doing payment for purchasing. For time saving and various offers on credit card and debit card customer prefer on line shopping like various platform Amazon, Flip cart, big basket etc. For online transaction security is prime concern. There is various type of attack possible during online transaction, stealing of password, fraud transaction, and meet in middle attack etc. During the online transaction stealing confidential information like OTP, transfer money from someone account to another account is a crime. In the digital world fraud during the online transaction day by day increases exponentially. To detect the unauthenticated transaction and fraud during online used various methods. Data is playing very important role during the online fraud. So, knowledge discovery is most frequently used to protect online fraud. In this paper suggested a technique based on knowledge discovery and machine learning methods, we strive to develop the best model possible in this research study to predict transactions involving fraud and transactions involving no fraud. Fraud detection uses a variety of machine learning techniques, including K-Means clustering methods, Support Vector Classifier, Logistic Regression, and Anomaly Detection Algorithm Techniques. After analysis it was found that Anomaly Detection Algorithm Techniques gives best accuracy for fraud detection 99.85%.

In recent times, AI has become so advanced that it can assist individuals in solving early medical problems by simply bringing up symptoms and medical images. In this work, we developed a web prototype named UPCHAAR-AI Healthcare to experiment on these cool features.The prototype combines a chat interface and a deep-learning pipeline that scans X-ray images. We fine-tuned a pre-trained DenseNet121 on the public chest X-ray sets and obtained a score of approximately 96.1 percent accuracy in the test data. The chat application, which was developed using LangChain, allows users to enter how they feel and receive an easy-to-read summary. This is not a system to make official diagnoses, but to increase basic health awareness. We believe that combining conversational logic and automated image recognition can make simple health data more available, particularly when a physician is not available right away.

The paper describes ways in which machine learning (ML)-based intrusion detection systems (IDS) have utilized pattern recognition methods to find potential malicious behaviour in internet-of-things (IoT) networks. However, many ML-based IDSs are unable to help analysts understand what was found and what will be the recommend actions after identifying possible malicious activity. Thus, this research will assess the functionality of present-day IDS models within IoTenvironments and create a new cognitive-security model utilizing retrieval-augmented generation (RAG) to enhance the capabilities of an IDS system, allowing systems to gather relevant data to analyse and report findings and recommended actions to analysts. Our primary objective is to assist in the utilization of ML-based IDS through the creation of a RAGenhanced cognitive security framework for IoT. Our RAGenhanced framework will aid in the provision of relevant context and meaningful explanations to analysts during their responses to potential cyber threats in IoT networks. We believe that by reducing the analyst's mental workload when reviewing raw output, and providing them with relevant contextual information, our enhanced cognitive security framework will allow teams to respond quickly, clearly and confidently to cyberattacks occurring within IoT-networks. We will accomplish our objectives by performing a literature review of existing studies regarding the performance of ML-based IDS models in IoT environments. Additionally, we will perform experimentation to evaluate the effectiveness of a RAG-enhanced cognitive security model for IoT. These experiments will consist of evaluating the ability of our proposed model to discover previously unknown attacks and determining if our model can reduce the time analysts spend resolving discovered attacks.

Mobile Agents are a new type of computing that is replacing the client-server approach. Mobile agents are little pieces of code that function automatically on behalf of the owner. Many applications, such as e-commerce, parallel computing, network management, and health care, use mobile agents. The healthcare industry is one of the most growing fields in any country. As the population increases day by day the requirement of medical resources is proportionally increasing. Due to high patient demand and a severe lack of medical resources, a remote medical healthcare system is required. However, the deployment of remote healthcare systems over the Internet introduces a new set of challenges, including interoperability among heterogeneous networks and the need to navigate through multiple public systems dispersed over insecure networks. This paper explores how mobile agents can effectively tackle these challenges, especially in heterogeneous and potentially malicious environments. A key focus of this research is the development of a mathematical model for secure medical information retrieval. This model incorporates a variable threshold secret-sharing mechanism, employing the Chinese remainder theorem and multiplicative inverse with modular arithmetic at different levels. By integrating these cryptographic techniques, the proposed approach ensures the confidentiality and integrity of medical information during its retrieval, contributing to the overall safety and robustness of mobile agent computing in healthcare scenarios.

Water quality is an essential measure for maintaining health and quality of life. In this study, the water quality index has been computed for Gautam Buddha Nagar using the Brown et al. method by collection of 51 groundwater samples. For this purpose, six physicochemical water parameters were analysed, namely pH, Hardness, Turbidity, C.O.D., D.O. and B.O.D. The readings indicate that the groundwater condition of Gautam Buddha Nagar is extremely poor. The computation of the Water Quality Index is a complex task. The study found that when all input variables are available, Machine Learning Techniques can be employed to vastly reduce the complexity in the computation while giving a high accuracy of 99.99% using Linear Regression, followed by an accuracy of 99.97% using Support Vector Regressor. Collection of input data is a time-consuming and costly process, therefore, the dimensionality of the input data was reduced through correlation analysis in an attempt to compute the water quality index by using just a single parameter. The best score of 81.05% was obtained using Linear Regression when Turbidity was used as the only feature, due to its high correlation with the target variable. The algorithms used for this analysis are: Linear Regression, Support Vector Regressor, Decision Tree and Random Forest Regressor.

This paper explores the application of deep learning techniques to heal online conversations on social media platforms. Online comment sections are public spaces but are often toxic with hate speech, personal attacks and offensive language. This paper addresses the challenge of healthy communication in digital spaces. This review consists of comparative analysis of various deep learning models like BERT, GPT, SVM and RNN model– LSTM for natural language processing techniques to classify comments based on their toxicity levels. It is observed that BERT performs better among all with an accuracy of 92.4%. This helps to promote respectful online conversations. A reliable & dependable way to deal with comment toxicity

Illegal fishing is a pervasive and destructive global issue that poses a significant threat to maritime ecosystems and the resilience of fisheries. Illegal, unregulated, and unreported (IUU) fishing leads to the extinction of the fishing population. Many researchers have presented various approaches to detect illegal fishing, for example, using sensors, image recognition, and convolutional neural networks (CNNs) but each one has some limitations. Our research aims to compare different vessel gear types to select the best vessel container that can be easily monitored and less prone to illegal activities. To achieve this, our research proposed an optimization method that involves hyperparameter selection using a genetic algorithm instead of a grid search. Using the crossover method of the genetic algorithm our model is compatible with larger datasets and unknown search space which is not possible in the baseline algorithm i.e. grid search. Moreover, after applying the genetic hyperparameter optimization technique, the overall accuracy, recall, and F1 score is increased for all vessel types significantly. While comparing our optimized model with the existing model with different evaluation metrics, our model’s performance is outstanding.

Aspect-based opinion mining (ABOM) has emerged as a crucial task in natural language processing (NLP) aimed at extracting fine-grained opinions about specific aspects of products, services or entities from textual data. With the exponential growth of user-generated content on the Internet, such as product reviews, social media posts and forum discussions, understanding the opinions expressed towards different aspects has become increasingly important for businesses, researchers and consumers alike. Traditional sentiment analysis approaches often treat the entire document or sentence as a unit, neglecting the fact that opinions are often expressed towards specific aspects or features. However, ABOM goes beyond overall sentiment analysis by identifying the aspects being discussed and extracting opinions related to each aspect individually. This granular analysis provides deeper insights into the strengths and weaknesses of products or services, enabling businesses to make informed decisions and enhancing user experiences. In natural language processing, aspect-based opinion mining is an essential task that seeks to analyse opinions stated in text documents about particular qualities or aspects of entities, products or services. Organisations and customers may make better decisions with the aid of the customer-based summary produced from the identified aspect words. The main goal of this work is to estimate the polarity of aspect terms in a given textual collection. Sentiment polarity estimation is needed for a large number of samples or reviews in the dataset. In this paper, the approach of deep memory network is utilised. When assessing the sentiment polarity of an aspect, the deep memory network technique explicitly takes into account the significance of each context word. Deep memory networks are able to store and make use of historical context, which facilitates a more sophisticated comprehension of the connections among various aspects and opinions in the reviews. They also enhance the recognition and categorisation of emotions linked to particular features. Experiments are conducted on laptop and restaurant datasets and the performance of the model is then evaluated with different classifiers.

In Natural Language Processing (NLP), Sentiment Analysis (SA) is a fundamental process which predicts the sentiment expressed in sentences. In contrast to conventional sentiment analysis, Aspect-Based Sentiment Analysis (ABSA) employs a more nuanced approach to assess the sentiment of individual aspects or components within a document or sentence. Its objective is to identify the sentiment polarity, such as positive, neutral, or negative, associated with particular elements disclosed within a sentence. This research introduces a novel sentiment analysis technique that proves to be more efficient in sentiment analysis compared to current methods. The suggested sentiment analysis method undergoes three key phases: 1. Pre-processing 2. Extraction of aspect sentiment and 3. Sentiment analysis classification. The input text data undergoes pre-processing through the implementation of four typical text normalization techniques, which include stemming, stop word elimination, lemmatization, and tokenization. By employing these methods, the provided text data is prepared and fed into the aspect sentiment extraction phase. During the aspect sentiment extraction phase, features are obtained through a series of steps, including enhanced ATE (Aspect Term Extraction), assessment of word length, and determination of cosine similarity. By following these steps, the relevant features are extracted on the basis of aspects and sentiments involved in the text data. Further, a hybrid classification model is proposed to classify sentiments. In this work, two of the Deep Learning (DL) classifiers, Bi-directional Gated Recurrent Unit (Bi-GRU) and Long Short-Term memory (LSTM) are used in proposing a hybrid classification model which classifies the sentiments effectively and provides accurate final predicted results. Moreover, the performance of proposed sentiment analysis technique is analyzed experimentally to show its efficacy over other models.

Now a days, the frequency of mental illness particularly depression has alarmingly increased. Inadequate early intervention and support for depression identification have led to the rise in associated disorders like anxiety, bipolar, and sleep disorders, as well as, in extreme situations, self-harm and suicide. It is extremely difficult to identify people with mental health issues and to provide prompt therapies. In this paper, the literature review of various techniques is done to detect depression in the patients using machine learning and we found that the accuracy of existing depression diagnosis techniques, which rely on PHQ scores and patient interviews are inadequate. This paper presents the previous research methods to detect depression with advantages and disadvantages of each. It is observed that the detection using LSTM technique gives better results than other machine learning algorithms.

Understanding the differences of high accuracy and data driven customer behaviour is one of the essential components of success in the e-commerce industry because customer behaviour varies from person to person depending on their segmentations. Owners will be able to recognise their desired customers by comprehending customer behaviour. They will be able to better target their marketing efforts, boost sales, and control costs. Applications of artificial intelligence in this area have a significant positive effect on operations. The most attributes that can influence a customer’s behaviour can be found and predicted by business partners using a data mining prediction model. The current study therefore identifies better ways to improve business decision-making through the use of AI and data analytics, which will aid in comprehending customer behaviour. This research, brings forth ideas and concepts to make models more data driven and more accurate through a more reactive approach on model designing, through clever infrastructural designing. This paper proves that the proposed method can assure users to get better results through reactive-data methodologies. This paper proves it practically by taking a simple classification problem, in contract to customer-behaviour prediction.

Mining the sentiment target included in a sentence or text is the main aim of Aspect Based Sentiment Analysis (ABSA). This task's main challenge is to efficient extraction of a specific sentiment item's sentiment polarity. This work proposes a model namely Improved ABSA with Deep Belief Network-Recurrent Neural Network (DBN-RNN), which includes 3 working phases. Processes like stemming, stop word removal, lemmatization as well and tokenization are conducted in the initial pre-processing phase. Furthermore, in the aspect sentiment extraction phase, improved aspect term extraction (I-ATE) along with cosine similarity and word co-occurrence are used to extract the complex features from the pre-processed data. In the sentiment analysis phase, a hybrid classification model named DBN-RNN is utilized to effectively categorize the sentiments as neutral, positive, and negative polarities. The performance of proposed work is evaluated in terms of different performance measures.

Moreover half of the population of India relies on agriculture for a living, making it the foundation of the nation’s economy. Agriculture’s future viability is now being threatened by weather, temperature, and other environmental variables. One use of machine learning (ML) is the Crop Yield Prediction (CYP) decision support tool, which provides suggestions about which crops to cultivate and what to perform during the crop’s growth season. Multi-source data for soils, climates, and remotely sensed vegetation indices particular to each site are needed for yield prediction. It is difficult to cope with model uncertainty when using complicated data-model fusion algorithms for crop growth monitoring and yield prediction Several aspects must be considered while developing an accurate and effective model for agricultural yield estimation depending on climate, crop illness, crop classification based on development phase, and other considerations, several research proposals for agricultural development have been made. This study explores severalML techniques for estimating agricultural yields and offers a thorough evaluation of the effectiveness of the methods and we found that the accuracy with Random Forest is higher i.e. 99.31% among all.

Mobile agent is a piece of computer code that organically goes from one host to the another in a consistent or inconsistent environment to distribute data among users.An autonomous mobile agent is an operational programme that may migrate from one computer to machine in different networks under its own direction.Numerous health care procedures use the mobile agent concept.An agent can choose to either follow a predetermined course on the network or determine its own path using information gathered from the network.Security concerns are the main issue with mobile agents.Agent servers that provide the agents with a setting for prosecution are vulnerable to attack by cunning agents.In the same way agent could be carrying sensitive information like credit card details, national level security message, passwords and attackers can access these files by acting as a middle man.In this paper, optimized approach is provided to encrypt the data carried by mobile agent with Advanced Encryption Standard (AES) algorithm and secure key to be utilized by the AES Encryption algorithm is generated with the help of Hopfield Neural Network (HNN).To validate our approach, the comparison is done and found that the time taken to generate the key using HNN is 1101ms for 1000 iterations which is lesser than the existing models that are Recurrent Neural Networks and Multilayer Perceptron Network models.To add an additional level of security, data is encoded using hash maps which make the data not easily readable even after decrypting the information.In this way it is ensured that, when the confidential data is transmitted between the sender and the receiver, no one can regenerate the message as there is no exchange of key involved in the process.