Despite the continuous growth and the widespread support of renewable energy sources, solar and wind power plants pose new challenges for Transmission System Operators and Distribution System Operators. Their uncontrollability limits their applicability; therefore, to encourage their further growth, fundamental modifications are needed. The research presented in this paper focuses on the predictive control of storage-based renewable power plants, and suggests a new model for profit optimization. Profit optimization is based on electricity price prediction and effective trading strategies that match the projected electricity prices. For the electricity price prediction, a recurrent Long Short-Term Memory neural network was developed and fine-tuned. For the optimization of the electricity trading, two trading strategies, namely an adaptive gradient-descent method and a differential evolution method were developed. Both optimization techniques were tested on mathematical models of most commercially available hybrid inverter systems and one year of historical data of electricity prices. As a result, a novel model predictive control workflow and sizing guide is proposed, which may significantly increase the profit generated by the system.
The huge amount of data stored in healthcare databases allows wide range possibilities for data analysis. In this article, we present a novel multilevel analysis methodology to generate and analyze sequential healthcare treatment events. The event sequences can be generated on different abstraction levels automatically from the source data, and so they describe the treatment of patients on different levels of detail. To present applicability of the proposed methodology, we introduce a short case study as well, in which some analysis results are presented arising from the analysis of a group of patients suffering from colorectal cancer.
During the training of neural networks, selecting the right stopping criterion is crucial to prevent overfitting and conserve computing power. While the early stopping and the maximum number of epochs stopping methods are simple to implement, they have limitations in identifying the point during training where the training and the validation loss start to diverge. To overcome these limitations, we propose a general correlation-based stopping criterion called the Correlation-Driven Stopping Criterion (CDSC). The CDSC stops the training process when the rolling Pearson correlation of the loss metrics between the training and validation datasets decreases below a pre-defined threshold. To show the effectiveness of the newly proposed Correlation-Driven Stopping Criterion, its effectiveness was compared with the effectiveness of the early stopping and the maximum number of epochs stopping methods across multiple common machine learning problems and neural network models. Our study shows that the proposed Correlation-Driven Stopping Criterion can enhance the out-of-sample performance of all tested neural network models while conserving computing power.
Retrospective studies suffer from drawbacks such as selection bias. As the selection of the control group has a significant impact on the evaluation of the results, it is very important to find the proper method to generate the most appropriate control group. In this paper we suggest two nearest neighbors based control group selection methods that aim to achieve good matching between the individuals of case and control groups. The effectiveness of the proposed methods is evaluated by runtime and accuracy tests and the results are compared to the classical stratified sampling method.
Case-control studies rely on the fact, that individuals of the case group are similar to the individuals of the control group, except for the feature under investigation. However, measuring the similarity of the case group and the control group is a complex task. The widely applied statistical methods compare the distributions of the characteristic features individually or use dimensionality reduction methods to estimate the similarity of the two groups. In this paper, three complex dissimilarity measures are proposed for the evaluation of the degree of similarity of case and control groups. These measures take all characteristic features of individuals into consideration at the same time and do not apply dimensionality reduction. The applicability of the proposed measures is tested and presented by Monte Carlo simulations.
