Search

Misdiagnosis is a critical issue in healthcare, which can lead to severe consequences for patients, including delayed or inappropriate treatment, unnecessary procedures, psychological distress, financial burden, and legal implications. To mitigate this issue, we propose using deep learning algorithms to improve diagnostic accuracy. However, building accurate deep learning models for medical diagnosis requires substantial amounts of high-quality data, which can be challenging for individual healthcare sectors or organizations to acquire. Therefore, combining data from multiple sources to create a diverse dataset for efficient training is needed. However, sharing medical data between different healthcare sectors can be problematic from a security standpoint due to sensitive information and privacy laws. To address these challenges, we propose using blockchain technology to provide a secure, decentralized, and privacy-respecting way to share locally trained deep learning models instead of the data itself. Our proposed method of model ensembling, which combines the weights of several local deep learning models to build a single global model, that enables accurate diagnosis of complex medical conditions across multiple locations while preserving patient privacy and data security. Our research demonstrates the effectiveness of this approach in accurately diagnosing three diseases (breast cancer, lung cancer, and diabetes) with high accuracy rates, surpassing the accuracy of local models and building a multi-diagnosis application. [ABSTRACT FROM AUTHOR]

User-generated content on social media platforms has reached big data levels. Sentiment analysis of this data provides opportunities to gain valuable insights into any domain. However, analyzing real-world data may confront the challenge of class imbalance, which can adversely affect the generalization ability of models due to majority class overfitting. Therefore, having an efficient model that manages any scenario of imbalanced data is practically needed. In this light, this work proposes different models based on studying the impact of data quality and transfer learning through pre-trained embeddings on boosting minority class detection. The proposed models are tested on imbalanced datasets related to social media and education. The experimental results highlight the effectiveness of Wor2vec, Glove, and Fasttext embeddings with preprocessed data. In contrast, BERT embeddings present better results with no-preprocessed data. Furthermore, in comparison with other methods, the best-performing model resulting from this study shows outperformance with notable improvements. [ABSTRACT FROM AUTHOR]

Nowadays, every kind of devices with different sizes and shapes, from lamp to kitchen appliances and industrial machines, are connected and shares information digitally in large scale. Despite this tendency to use Internet in such gadgets, vast amounts of data are generated creating new challenges for researchers to analyze and control them. On the other side, Deep Learning (DL) is an appropriate tool for dealing with Internet of Things (IoT) needs, such as analyzing data, making predictions, classifying data. Acquiring the most accurate neural network inside a sensible run-time is a challenge. However, metaheuristics are the key to the success of the application of DL on IoT big data due to non-deterministic polynomial time (NP hard) problems in these areas. Many papers were published about metaheuristic in optimizing deep leaning models, but the literature lacks a study that precisely investigate the relationship between IoT, deep learning and metaheuristic. In this paper, a review of the metaheuristic's usages in the realm of IoT are presented. [ABSTRACT FROM AUTHOR]