Alba Márquez-Rodríguez - GrunCrow

Tidal energy represents a clean and sustainable source of energy generation that can address renewable energy challenges, especially the global challenge of optimizing alternatives for stable supply. Although tidal stream energy extraction technology is in the early stages of development, it shows great potential compared to other renewable energy sources. The main objective of this research is to provide a digital tool for the optimization of the installation of turbines through fuzzy logic. The methodology in this study includes the design and development of a fuzzy-logic-based tool for this purpose. Design criteria included parameters such as salinity, temperature, currents, depth, and water viscosity, which affect the performance of tidal turbines. These parameters are obtained from the geographic location of the installation. A decision-making system is provided to support the tool. The designed fuzzy logic system evaluates the suitability of different turbine locations and presents the results through graphics and probability of success percentages. The results indicate that currents and temperatures are the most limiting factors in terms of potential turbine locations. The program provides a practical and efficient tool for optimizing the selection of tidal turbines and generating energy from ocean currents. This tool is evaluated and validated through different cases. With this approach, the aim is to encourage the development of tidal energy and its adoption worldwide.

This paper introduces a framework for feature selection. Firstly, it establishes a percentage in order to set a cut-off point in the context of feature ranking based on feature selection. Secondly, it combines a couple of feature ranking methods using a common percentage from those features. This combination has a weak to a strong influence in the class label that is run in an independent way and merges the solutions achieved by every ranking-based feature selection. The proposed methodology has been called Feature Ranking merging (FRmgg). It has been tested with three high-dimensionality binary classification data sets, which have been assessed with three classifiers and two measures have been reported. The results are acceptable since the feature space reduction is convenient compared to the raw data set and the results in some cases are better.

Data mining (DM) and Analytics may take advantage of including any data preparation procedure from the feature perspective. DM is often used to analyze and extract useful patterns and information from datasets for prediction and decision making. The stage of data pre-processing in Machine Learning can also be demanding in the cost of any data-driven model [1]. It is an essential part that have to be done before the creation of any model and prediction.

Depending on different dataset requirements, machine learning techniques can be categorized as the taxonomy in Figure 1 shows [15]. Supervised learning requires large amount of labelled data as input and also as a method of evaluating model accuracy when output is produced, hence the term ‘supervised’. Unsupervised learning on the other hand refers to the model’s manipulation of unlabeled data including clustering and reducing dimension, therefore discovering hidden patterns without human intervention, hence the term ‘unsupervised’.

This paper proposes a methodology to feature sorting as well as feature selection in the context of supervised machine learning algorithms. Feature sorting has been revealed as a step which may play a paramount role in machine learning. Nonetheless, the scalability is an important drawback. This paper proposes to add a further stage in order to only retain attributes with a positive influence (att+) and limiting them in a predefined percentage of att+ set. This contribution aims at introducing a new methodology where all attributes are not included in the data mining task but also the positive influence ones till a certain limit. We have followed two different types of sorting by means of different feature ranking methods. The approach has been assessed in three binary problems with a number of features between 1000 and 10000, and a number of instances from 200 to 7000; the test-bed includes challenging data sets from NIPS 2003. According to the experimental results for InfoGain and GainRatio the 90% of the attributes with positive influence are enough to get results in most of the cases comparable to the results with raw data taking into account that the required time to train the classifiers is shorter and hence in the non-required time we may be able to process more instances.