Search

Most research on movie recommender systems has been conducted with Collaborative Filtering (CF) methods. The lack of sufficient information about users' interests in bootstrapping is one of the most critical problems of the CF method. Using a content-based filtering method can mitigate some of these problems. On the other hand, recent research has proven that utilising contextual information about the movie, such as genre, actors, and cast, can increase the efficiency of the recommender system. This paper uses a combined Genetic Algorithm (GA) and content-based filtering to find the best combination of genre, cast, and crew weights. We first convert each movie's contextual information into a descriptive vector, including cast, crew, and genre. Then, we calculate the distance between each pair of sentence vectors using two separate approaches, fully connected and metadata-based. Finally, with GA, we tune the weight of each of the contextual information of movies to maximise the recommender system's efficiency. Performance evaluation on the well-known MovieLens dataset shows that GA can improve the Precision@k criterion by producing fewer, more accurate recommendations. Weight adjustment by GA improves the F-Measure metric by approximately 58%. This, in turn, can improve Precision and Recall metrics. Also, the GA offers a higher correct recommendation rate than other methods. [ABSTRACT FROM AUTHOR]

Edge computing allows IoT tasks to be processed by devices with passive processing capacity at the network's edge and near IoT devices instead of being sent to cloud servers. However, 5G‐enabled architectures such as Fog Radio Access Network (F‐RAN) use smart devices to bring the delay down to even a few milliseconds. This is important, especially in latency‐sensitive applications such as online digital games. However, a trade‐off must be made between the delay and energy consumption. If too many tasks are processed locally on edge devices or fog servers, energy consumption increases because mobile devices such as smartphones and tablets have limited energy charges. This paper proposes a Deep Reinforcement Learning (DRL) method for offloading optimization. In designing states, we consider all three critical components of memory consumption, the number of CPU cycles, and network mode. This makes the modeling aware of the workload of the tasks. As a result, the model matches the requirements of the real world. For each mobile device that submits a task to the system, we consider a reward. It includes the total delay of tasks and energy consumption. The output of our DRL model specifies to which edge/fog/cloud device each task should be offloaded. The results show that the DRL technique produces less resource waste than RL when the number of tasks is very high. In addition, DRL consumes 30% less network resources than the FIFO method. As a result, DRL provides a better trade‐off between offloading and local execution than other methods. [ABSTRACT FROM AUTHOR]

Summary: One of the practical preferences of cloud service providers is to use specialized physical hosts. In other words, the goal is to place homogeneous virtual machines (VMs) on the physical host according to performance criteria such as energy consumption, resource wastage, and utilization. virtual machine placement (VMP) falls into NP‐hard knapsack problems. To overcome the time complexity, the use of heuristic and metaheuristic methods has attracted the attention of researchers. In this paper, we use an entropy‐based method for VMP for the first time. The proposed method tries to place the VMs on physical machines by considering the type of VMs to minimize entropy. Entropy is a measurable property that is more associated with disorder, randomness, or uncertainty. We use one of the most common entropy criteria called the Gini coefficient. In summary, among the different placement combinations of VMs, those that can minimize the Gini coefficient are preferred. We then solve the multi‐objective problem with the non‐dominated sorting genetic algorithm (NSGA‐III). We also combine this method with differential evolution methods to improve the quality of solutions. Recent research in other engineering fields has shown that combining metaheuristic methods with differential evolution methods increases the rate of convergence toward the optimal solution. The simulation results on the CloudSim simulator, along with statistical analysis, show that the entropy‐based method has a significant improvement over the state‐of‐the‐art methods in terms of significant performance criteria such as utilization, resource wastage, and energy consumption. [ABSTRACT FROM AUTHOR]

The spot market is one of the most common cloud markets where cloud providers, such as Amazon EC2, rent their surplus computing resources at lower prices in the form of spot virtual machines (SVMs). In this market, which is often managed through an auction mechanism, users seek optimal bidding strategies for renting SVMs to minimize cost and risk. Uncertainty in the price of SVMs and their low availability/reliability is a challenging issue to bid on the user side. In this paper, we present a robust model for minimizing the cost of executing tasks by considering the uncertainty of the price of SVMs based on the Information Gap Decision Theory (IGDT). It evaluates the risk-aversion and the risk-seeker nature of the user's bidding strategy and measures the cost of risk/immunity. The main advantage of this method is that it formulates user decisions without the need for any presumption about the price distribution of SVMs. With this decision-support system, users can rely on the predicted confidence intervals to make the optimal bid for future time slots according to the selected risk level. The results are compared with Monte Carlo simulations and a scenario-based approach to evaluate the effectiveness of the proposed IGDT-based model. Evaluation results based on historical Amazon EC2 prices confirm the efficiency of the proposed method to handle the uncertain nature of the price of SVMs in terms of significant criteria such as robustness cost, opportunity cost, uncertainty budget, and execution time. [ABSTRACT FROM AUTHOR]