BACKGROUND: No-shows, a major issue for healthcare centers, can be quite costly and disruptive. Capacity is wasted and expensive resources are underutilized. Numerous studies have shown th...

This research addresses a variant of the traveling salesman problem in drone-based delivery systems known as the TSP-D. The TSP-D is a combinatorial optimization problem in which a truck and a drone collaborate to deliver parcels to customers, with the objective of minimizing the total delivery time. Determining the optimal solution is NP-hard; thus, the size of the problems that can be solved optimally is limited. Therefore, metaheuristics are used to solve the problem. Metaheuristics are adaptive and intelligent algorithms that have proved their success in many similar problems. In this study, a solution to the TSP-D problem using the greedy, randomized adaptive search procedure with two local search alternatives and a self-adaptive neighborhood selection scheme is presented. The proposed approach was tested on 200 instances with different properties from the publicly available “Instances of TSP with Drone” benchmark. Results were evaluated against state-of-the-art algorithms. Non-parametric statistical tests concluded that the proposed approach has comparable performance to the rival algorithms ($p=0.074$) in terms of tour duration. The proposed approach has better or similar performance in instances where the drone and truck have the same speed ($\alpha =1$).

Data preprocessing has a profound effect on the performance of the learner. Before attempting medical data classification, characteristics of medical datasets, including noise, incompleteness, and the existence of multiple and possibly irrelevant features, need to be addressed. In this paper, we show that selecting the right combination of preprocessing methods has a considerable impact on the classification potential of a dataset. The preprocessing operations considered include the discretization of numeric attributes, the selection of attribute subset(s), and the handling of missing values. The classification is performed by an ant colony optimization algorithm as a case study. Experimental results on 25 real-world medical datasets show that a significant relative improvement in predictive accuracy, exceeding 60% in some cases, is obtained

Human activity recognition (HAR) is a crucial task in various applications, including healthcare, fitness, and the military. Deep learning models have revolutionized HAR, however, their computational complexity, particularly those involving BiLSTMs, poses significant challenges for deployment on resource-constrained devices like smartphones. While BiLSTMs effectively capture long-term dependencies by processing inputs bidirectionally, their high parameter count and computational demands hinder practical applications in real-time HAR. This study investigates the approximation of the computationally intensive BiLSTM component in a HAR model by using a combination of alternative model components and data flipping augmentation. The proposed modifications to an existing hybrid model architecture replace the BiLSTM with standard and residual LSTM, along with convolutional networks, supplemented by data flipping augmentation to replicate the context awareness typically provided by BiLSTM networks. The results demonstrate that the residual LSTM (ResLSTM) model achieves superior performance while maintaining a lower computational complexity compared to the traditional BiLSTM model. Specifically, on the UCI-HAR dataset, the ResLSTM model attains an accuracy of 96.34% with 576,702 parameters, outperforming the BiLSTM model’s accuracy of 95.22% with 849,534 parameters. On the WISDM dataset, the ResLSTM achieves an accuracy of 97.20% with 192,238 parameters, compared to the BiLSTM’s 97.23% accuracy with 283,182 parameters, demonstrating a more efficient architecture with minimal performance trade-off. For the KU-HAR dataset, the ResLSTM model achieves an accuracy of 97.05% with 386,038 parameters, showing comparable performance to the BiLSTM model’s 98.63% accuracy with 569,462 parameters, but with significantly fewer parameters.