Skin melanoma cases have increased significantly in recent years, partly due to environmental changes. Early detection of skin melanoma helps patients receive timely treatment. While deep learning methods have been used for melanoma detection, but there is still scope for improvement. In this study, a combined approach based on three different deep learning techniques is developed to detect skin melanoma more effectively. The skin images are first processed using a black-hat transformation to remove hair. Both individual deep learning methods and an ensemble of neural networks are applied for melanoma detection. The method’s performance is evaluated using metrics like accuracy, recall, precision, and F1-score. The proposed ensemble method shows better results compared to existing approaches, achieving an accuracy of 98.52

Storytelling is an important framework for entertainment, sharing culture, and emotion; yet, AI techniques still struggle with some challenges: limited controllability and resource-intensive deployment. This research demonstrates a multi-objective-based storytelling framework that generates narratives conditioned on age group, genre, emotional tone, and paragraph length. Built on LLaMA v3-8B with Quantized Low-Rank Adapter (QLoRA) fine-tuning, the framework updates fewer than 1% of parameters while preserving efficiency. The proposed framework extends the DesigningPrompts corpus with ~31K labelled interesting stories using open-source techniques and heuristics, enabling modern conditioning. A multi-stage strategy—baseline fine-tuning, instruction-tuned prompting, and hybrid outline–expansion with End-of-Sequence (EOS) enforcement—enhances structural completeness and controllability. Stage 3 reduces incomplete results by 40% and boosts human-rated coherence by 25% over the baseline model. This deployment uses a hybrid edge-cloud framework: inference runs server-side, while a Raspberry Pi hardware gives a lightweight UI and real-time narration. The outcomes demonstrate that parameter-efficient fine–tuning, integrated with sensor deployment, enables cost-effective storytelling for education and infotainment-based applications.

Depression is a mental health disorder that profoundly affects how a person thinks, feels, and behaves. Depression is a common and serious problem today, and prevention is essential. In today’s digital age, people freely express their feelings on social media. Therefore, depression can be detected by analyzing social media text and emotions using machine learning techniques. The purpose of this paper is to detect depression from Reddit posts. Natural Language Processing (NLP) and Ensemble Learning have been used on a Reddit dataset of 7731 posts to achieve this purpose. Basic preprocessing techniques such as data cleaning, tokenization, stemming, and lemmatization are applied while ensuring that pronouns were preserved due to their linguistic significance. Term FrequencyInverse Document Frequency (TF-IDF) has been applied for feature extraction as well as vectorization. A Stacking Ensemble model is proposed, integrating XGBoost, Support Vector Machine (SVM), and Multinomial Naive Bayes (MNB) as base learners and Logistic Regression (LR) is a meta-learner. The model achieved Accuracy 97.09%, Precision 98.13%, Recall 95.95% and F1-score 97.03% which was better than individual classifiers. The results prove that combining various Machine Learning techniques through stacking significantly reduces both variance and bias, leading to more accurate, stable and consistent predictions. Index Terms-Depression Detection, Ensemble Learning, Machine Learning, Natural Language Processing, Social Media.

Chronic kidney disease (CKD) is a lifelong, dangerous disorder caused by either kidney cancer or decreased kidney function. The progression of this chronic illness can be stopped or slowed down to an end stage where the only ways to save a patient’s life are dialysis or surgery. Therefore, early-stage identification of kidney disease is necessary. Patients who are aware of potential risk factors will be able to recognize the dangers of CKD and receive treatment early. Chronic kidney impairment is more common in older adults, those with diabetes, and those with high blood pressure. In order to conquer these obstacles, methods based on machine learning have been employed to quickly diagnose diseases. Stochastic Gradient Boosting (SGB), XgBoost, Cat Boost Classifier, Decision Tree Classifier, Random Forest Classifier, AdaBoost Classifier, Gradient Boosting Classifier, Extra Trees Classifier, and KNN Classifier are some of the machine learning techniques that have been utilized in this work. These algorithms were trained using data from the UCI repository. Among all the classifiers that have used in work, it has been found that Extra Trees Classifier (ETC) performs really well and got the highest accuracy of 99 percent.

The stock market is most dominant in today's world in determining the country's economic growth. Accurate price prediction is very necessary and important but faces various difficulties due to the uncertain and fluctuating nature of the stock market. Therefore, a strong prediction model is highly desirable for accurate price prediction. This paper focuses on the effective prediction of the stock market price using historical data and technical indicators. To obtain the desired result this paper proposes a novel multi-layer method. The first layer is based on the Variational Mode Decomposition technique, followed by the Optimized-Kernel Extreme Learning Machine in the second layer and Long Short Term Machine being the final layer. LSTM being less commonly applied in the field of financial market prediction. Hence an attempt has been made in this study for implementing the LSTM for price prediction. The price of different stock markets is used for experimental and validation purposes. Various single-stage methods are used for comparison purposes. The technical indicators along with the historical data are selected as the input variable for prediction purposes. The result analysis shows that the proposed hybrid model performs better than various other techniques mentioned in this study.

We assess the benefits of using simple and successive relaying in energy harvesting networks and compare the simple and successive relay networks in terms of maximum achievable throughput. Both the relay and source have the potential to harvest energy and do not generate any traffic on their own. This study investigates the interplay between energy harvesting, relaying, and stability. We assume that the source transmits to the destination using relay-based network cooperation. The relay regulates the source arrival data packets by accepting the part of the successfully received data packets received in the relay. However, our work is based on two relay networks. In the case of a simple relaying network, we exploit the partial relay cooperation in which flow can be controlled by partial relaying. The system should be stable when the relay and source are stable, and we applied the stability condition individually for both the relay and source. Then the stability for both the relay and the source are combined to get a stable system. The relay network secure communication depends on certificateless cryptography and the communication cost is very less. Our model work for small devices and provides the security authenticity with a trusted third party. We compare the successive partial relay network's approach performance with the full relay, no-relay, and simple relay performance.

Sarcasm is a type of communication that involves using words with meanings opposite to their literal definition to create humor or mock someone. This form of communication can be confusing as it often uses positive words to express negative feelings, making it difficult for people to understand the speaker’s intention. Detecting sarcasm in text can be challenging as it changes the polarity of the sentence and the difference between the words used and the way it is spoken. As a result, sarcasm detection in news, comments, or tweets on social media poses a challenge for researchers. In this research, various word-level features have been studied to detect sarcasm from three benchmark datasets, which include the creation of an N-gram probability dictionary, negation words, and PoS tags. Different machine learning and hybrid deep learning models have been examined and compared with handcrafted features and also with word embedding features. The results demonstrate a maximum accuracy of 87

Corporations getting bankrupt has been a severe issue for investors, businesses as well as ordinary individuals. Several research works have been conducted over the years to accurately predict bankruptcy, with the earliest works depending on the financial metrics of the company taken under consideration and the latest ones trying to predict bankruptcy from various kinds of data of organizations. However, the present works do not capture the dynamic nature of the business world and the possibility of a turnaround scenario. Hence, predictive models that are spatiotemporally aware when they predict a firm’s financial distress are needed. Considering this imminent problem, our work focuses on building a unique spatiotemporal context-aware bankruptcy prediction model that can predict bankruptcy with the help of daily news articles of a company or its related companies and key financial metrics to predict bankruptcy. Knowledge graphs were used to represent the vast amount of textual data. Their embeddings, along with the financial metrics, were used in the classification process. In the first stage, various machine learning algorithms were used for the financial metrics, while for the textual data or the embeddings, attention-based LSTM was used. Next, both were assembled together in the second stage to form the final predictive model, which has given an accuracy of 0.97 on the test set and an F1 score of 0.95. We hope our novel approach to this problem helps those who are uncertain about the future of any organization in predicting its bankruptcy beforehand and thereby timely decision making.

Extreme caution should be exercised while dealing with any form of skin cancer, but especially malignant melanoma. It’s also on the rise, especially among whites who spend a lot of time outdoors in the sun. Because early diagnosis of melanoma may be beneficial and curative, it is crucial that it be detected at an early stage to improve survival rates. Accurate automatic skin lesion segmentation is in high demand due to the rapid proliferation of skin cancer. While deep learning models like CNN were widely utilized to enable proper segmentation, current encoder-decoder designs based on compactly connected networks (DenseNet) and residual networks (ResNet) were applied for skin lesion tasks. Complex parameter settings, a lack of multi-scale data, and an absence of appropriate information in pre-trained features all have an effect on the performance of skin lesion segmentation. This research proposes a system for segmenting skin lesions utilizing the UNet and Residual UNET (ResUNet) architectures to solve these issues. CNNs, which consist of encoders and decoders, form the basis of these designs. The architecture employs UNet and ResUNet to guarantee high-quality lesion segmentation at all times. The proposed models are evaluated on ISIC2018 and HAM10000 lesion pictures. Accuracy, dice coefficient, Jaccard index, sensitivity, and specificity are used to assess the performance of the models. The models’ efficacy is evaluated in light of state-of-the-art approaches.

Opinion spam detection is a challenge for online review systems and social forum operators. Opinion spamming costs businesses and people money since it deceives customers as well as automated opinion mining and sentiment analysis systems by bestowing undeserved positive opinions on target firms and/or bestowing fake negative opinions on others. One popular detection approach is to model a review system as a network of users, products, and reviews, for example using review graph models. In this article, we study the effects of network scale on network-based review spammer detection models, specifically on the trust model and the SpammerRank model. We then evaluate both network models using two large publicly available review datasets, namely: the Amazon dataset (containing 6 million reviews by more than 2 million reviewers) and the UCSD dataset (containing over 82 million reviews by 21 million reviewers). It has been observed thatSpammerRank model provides a better scaling time for applications requiring reviewer indicators and in case of trust model distributions are flattening out indicating variance of reviews with respect to spamming. Detailed observations on the scaling effects of these models are reported in the result section.

The widespread use of social media and its development have offered a medium for the propagation of fake contents quickly among the masses. Fake contents frequently misguide individuals and lead to erroneous social judgments. Individuals and society have been harmed by the dissemination of low-quality news content on social media. In this paper, we have worked on a benchmark dataset of news content and proposed an approach comprising basic natural language processing techniques with different deep learning models for categorising content as real or fake. Different deep learning models employed are LSTM, bi-LSTM, LSTM and bi-LSTM with an attention mechanism. We compared the outcomes by using one hot word embedding and pre-trained GloVe technique. On benchmark LIAR dataset, the LSTM achieved a better accuracy of 67.2%, while the bi-LSTM with GloVe word embedding reached an accuracy of 67%. An accuracy of 98.22% is achieved using bi-LSTM and 97.98% using LSTM on Real-Fake dataset. Fake news can be a menace to society, so if it is detected early, harmony can be maintained in society and individuals can avoid being misled.

Mobile connectivity and smart devices are spreading worldwide. As a result, the use of mobile devices and applications is rising exponentially. Therefore, nowadays hackers target such smart devices to steal information and misuse it for malicious purposes. It becomes absolutely essential to protect sensitive information such as app. permissions, login credentials, browse history, media contents etc. from intruders. Security can be breached easily if smart techniques are not devised to safeguard mobile data. In this article, an attempt is made to classify the different types of malware and to protect the sensitive information on Android devices that significantly reduce network congestion and improve network throughput by increasing data transmission. The proposed hybrid approach consists of AdaBoost, random forest and deep learning methods jointly classify the sophisticated malware. The empirical results indicate that this achieves better classification and detection accuracy and is capable of identifying the potential threat more efficiently.

One of the primary goals of many defence applications is the detection and immediate response to human activities. Human activity detection has been proposed using various technologies, such as surveillance cameras and sensor-equipped wearable devices. However, these technologies are not widely adopted because of issues with accessibility, privacy, cost, and convenience. In recent years, the ability to identify and detect human actions based on the properties of wireless signals has been seriously considered. This paper proposes a CNN-based hybrid approach for detecting human activities using Wi-Fi sensors. This hybrid model combines Shallow CNN and SqueezeNet, which uses transfer learning to SVM classification model. This new model reduces the number of false negatives as much as possible. The designed hybrid model has an average of 98.67% accuracy for ten-fold cross-validation.

Most of the non-small cell lung cancer is clinically examined using CT/PET images. But an accurate diagnosis by the radiologist is difficult while classifying the type of non-small cell cancer, which may lead to misdiagnosis. Hence, a method is required to accurately identify different types of non-small cell lung cancers, such as adenocarcinoma and squamous cell carcinoma for providing proper treatment to patients. One of the practical and feasible solution is deep learning based method that has the ability to adapt and learn. However, most of the deep learning methods have complexity issues. Hence, some optimization is required to make the networks less complex. The objective of the work is to use less complex methods for classifying the non-small cell lung cancer. In this work, dense neural network (VGG-16 and Resnet-50) that has complex structures and sparse neural networks (inceptio v3) that are less complex are used. Deep learning methods are employed to obtain features from CT images and accurately classify non-small cell lung cancer. To evaluate the method, 60 adenocarcinoma patients and 60 squamous cell carcinoma patients are considered. The sensitivity, specificity, and accuracy of the Inception v3 network are found to be 96.66 %, 99.12 % and 98.29 % respectively. Observations indicate that the inception v3 model outperforms VGG-16 and ResNet-50. Also, the inception v3 network that is a sparse neural network has less computational overhead as compared to the other two networks. Sparse deep learning techniques may help radiologists accurately classify non-small cell lung cancer using CT images.

Purpose The objective of the proposed work is to identify the most commonly occurring non–small cell carcinoma types, such as adenocarcinoma and squamous cell carcinoma, within the human population. Another objective of the work is to reduce the false positive rate during the classification. Design/methodology/approach In this work, a hybrid method using convolutional neural networks (CNNs), extreme gradient boosting (XGBoost) and long-short-term memory networks (LSTMs) has been proposed to distinguish between lung adenocarcinoma and squamous cell carcinoma. To extract features from non–small cell lung carcinoma images, a three-layer convolution and three-layer max-pooling-based CNN is used. A few important features have been selected from the extracted features using the XGBoost algorithm as the optimal feature. Finally, LSTM has been used for the classification of carcinoma types. The accuracy of the proposed method is 99.57 per cent, and the false positive rate is 0.427 per cent. Findings The proposed CNN–XGBoost–LSTM hybrid method has significantly improved the results in distinguishing between adenocarcinoma and squamous cell carcinoma. The importance of the method can be outlined as follows: It has a very low false positive rate of 0.427 per cent. It has very high accuracy, i.e. 99.57 per cent. CNN-based features are providing accurate results in classifying lung carcinoma. It has the potential to serve as an assisting aid for doctors. Practical implications It can be used by doctors as a secondary tool for the analysis of non–small cell lung cancers. Social implications It can help rural doctors by sending the patients to specialized doctors for more analysis of lung cancer. Originality/value In this work, a hybrid method using CNN, XGBoost and LSTM has been proposed to distinguish between lung adenocarcinoma and squamous cell carcinoma. A three-layer convolution and three-layer max-pooling-based CNN is used to extract features from the non–small cell lung carcinoma images. A few important features have been selected from the extracted features using the XGBoost algorithm as the optimal feature. Finally, LSTM has been used for the classification of carcinoma types.

The mango industry faces substantial annual economic losses due to diseases and pests. Farmers encounter difficulties in differentiating between various mango diseases, as their symptoms can closely resemble each other and even occur simultaneously. By making it possible to quickly and accurately detect and identify mango leaf illnesses, this study aims to offer a remedy. To accomplish this, the research employs state-of-the-art Convolutional Neural Networks (CNNs), which are advanced learning algorithms capable of automatically extracting and understanding complex features from raw images. Specifically, the study utilizes EfficientNet, a model based on CNNs, to facilitate the early identification and classification of mango leaf diseases. Additionally, advanced data augmentation techniques such as rotation, reflection, translation, and scaling are applied to enhance the collected dataset. This enhanced data is then used to train the EfficientNet model, which achieves an amazing 98% accuracy rate in identifying and categorizing mango leaf diseases. These findings underscore the CNN method’s effectiveness in detecting a wide range of mango leaf diseases, positioning it as an invaluable and practical tool for both farmers and agricultural experts.

Network intrusion detection systems are crucial for preserving cybersecurity in the face of increasingly sophisticated cyberthreats that target international networks. Conventional approaches usually use signature-based methods, which are not very good at spotting new attacks like zero-day vulnerabilities. The current study explores machine learning-based methods to enhance NIDS's ability to detect known and unexpected threats. Our classifiers include Gaussian Naive Bayes, Bernoulli Naive Bayes, Logistic Regression, XGB Classifier, and K-Nearest Neighbors Classifier. For training and evaluation, we employ two popular datasets, UNSW-NB15 and NSL-KDD, which provide comprehensive representations of both malicious and benign traffic. Our results suggest that machine learning models may considerably boost intrusion detection's adaptability and accuracy, even though issues like feature selection, data imbalance, and processing efficiency need more research. The foundation of next-generation cybersecurity systems that can adjust to the constantly shifting threat landscape can be machine learning, as this study shows. NIDS can more precisely identify known and unknown attacks by using AIIML methods that learn from historical data to identify patterns and irregularities. This study examines a number of popular machine learning classifiers for intrusion detection using various datasets. The results shows that our study strengthens cybersecurity defense against different intrusion attacks.

Fraudulent activities associated with the credit card is a pertinent problem often occurring in a global level. The customers are losing their trust with the financial institutions and the financial institutions are in a difficult state to win the goodwill of customers. A substantial number of researchers show interest to work on fraud detection in order to develop an optimized method or model to identify the fraudulent activities that are happening in a regular and continuous form with the credit card in our everyday life. Genetic algorithm (GA) and the potential solution-based particle swarm optimization (PSO) are two optimization algorithms, which can be considered along with the neural network to analyze the possible fraudulent transactions. The optimization algorithms help to make the learning process faster and optimized with a superior and better predictive accuracy value. The PSO-based neural network has been trained thoroughly and performance values are compared with GA-based neural network, by increasing the number of iterations and the population or number of swarms. It has been observed that algorithm based on PSO gives an optimized result for fraudulent transaction detection.