Internet of Things (IoT) devices built on different processor architectures have increasingly become targets of ad- versarial attacks. The recent trends in cross-architecture malware cyber-attacks result in mass compromises in smart environments and IoT malware threat hunting poses a contemporary and complex challenge for researchers. This paper proposes a robust cross-architecture IoT malware threat hunting model based on advanced ensemble learning (MTHAEL) using a novel approach. Our unique MTHAEL model using stacked ensemble of hetero- geneous feature selection algorithms and state-of-the-art neural networks to learn different levels of semantic features demon- strates enhanced IoT malware detection than existing approaches. MTHAEL is the first of its kind that effectively optimizes recur- rent neural network (RNN) and convolutional neural network (CNN) with high classification accuracies and consistently low computational overheads on different IoT architectures. Our proposed MTHAEL is experimentally evaluated comprehensively with a large IoT cross-architecture dataset of 21,137 rare samples and is shown to achieve 99.98% classification accuracy for ARM architecture samples, surpassing prior related works. Cross- architecture bench-marking is also performed on the training of our MTHAEL model with different architectures such as ARM, Intel80386, MIPS, and MIPS+Intel80386 individually. Two different hardware architectures such as Raspberry Pi 4 and Core-i5

Nowadays, sentimental analysis of consumers’ review is becoming much crucial in the marketing world. It is not just giving ideas to the firms that how consumers like their product or service, but it would also help them make their service better. In this article, the statistical method identifies the relationship of many factors in consumer feedback. It introduces a deep-based learning method called DSC (deep sentiment classifier) to determine whether or not to recommend the reviewed product thoroughly. Our suggested method also investigates the effect sizes of the feedback, such as positives, negatives, and neutrals. We used the women’s clothing review dataset containing 22,642 records after preprocessing of the results. Experimental studies show that the recommendations are an excellent positive sentiment indicator. In comparison, ratings become fuzzy performance metrics in product reviews. The 10-fold cross-validation analysis shows that the recommended form has the top F1 score (93.56%) in the sentimental classification on average and the recommended classification (88.32%) on average. A comparative description of other classifiers focused on machine learning, for example, KNN, random forest, logistic regression, decision tree, support vector machine multilayer perceptron, and naïve Bayes, also demonstrates that DSC gives the best possible result. We have tested DSC on the dataset IMDB (Internet Video Database), which includes the sentiment of the 50,000 movie reviews (25000 for training and 25000 for testing). In comparison to other baseline methods, DSC obtained an excellent classification score for this experiment.