A5011933761- Time Series Analysis and Forecasting
- Complex Systems and Time Series Analysis
- Metaheuristic Optimization Algorithms Research
- Stock Market Forecasting Methods
- Hydrological Forecasting Using AI
- Anomaly Detection Techniques and Applications
- Blind Source Separation Techniques
- Energy Load and Power Forecasting
- Ocean Waves and Remote Sensing
- Financial Risk and Volatility Modeling
- Spectroscopy and Chemometric Analyses
- Oceanographic and Atmospheric Processes
- Machine Learning and ELM
- Neural Networks and Applications
- Risk and Portfolio Optimization
- Neurological disorders and treatments
- Ecosystem dynamics and resilience
- Teaching and Learning Programming
- Photovoltaic System Optimization Techniques
- Neural Networks and Reservoir Computing
- Machine Learning and Data Classification
- E-Learning and Knowledge Management
- Wave and Wind Energy Systems
- Rough Sets and Fuzzy Logic
- MicroRNA in disease regulation
Universidad Loyola Andalucía
2020-2024
Universidad Loyola
2021-2023
University of Córdoba
2015-2019
Time-series clustering is the process of grouping time series with respect to their similarity or characteristics. Previous approaches usually combine a specific distance measure for and standard method. However, these do not take different subsequences each into account, which can be used better compare time-series objects dataset. In this article, we propose novel technique consisting two stages. first step, least-squares polynomial segmentation procedure applied series, based on growing...
This paper proposes a novel methodology for recovering missing time series data, crucial task subsequent Machine Learning (ML) analyses. The is specifically applied to Significant Wave Height (SWH) in the field of marine engineering. proposed approach involves two phases. Firstly, SWH each buoy independently reconstructed using three transfer function models: regression-based, correlation-based, and distance-based. distance-based exhibits best overall performance. Secondly, Evolutionary...
Abstract Artificial Neural Networks (ANNs) have been used in a multitude of real-world applications given their predictive capabilities, and algorithms based on gradient descent, such as Backpropagation (BP) variants, are usually considered for optimisation. However, these shown to get stuck at local optima, they require cautious design the architecture model. This paper proposes novel memetic training method simultaneously learning ANNs structure weights Coral Reef Optimisation (CROs),...
Parkinson's disease is characterised by a decrease in the density of presynaptic dopamine transporters striatum. Frequently, corresponding diagnosis performed using qualitative analysis 3D-images obtained after administration [Formula: see text]I-ioflupane, considering binary classification problem (absence or existence disease). In this work, we propose new methodology for classifying kind images three classes depending on level severity image. To tackle problem, use an ordinal classifier...
Randomized-based Feedforward Neural Networks approach regression and classification (binary multi-class) problems by minimizing the same optimization problem. Specifically, model parameters are determined through ridge estimator of patterns projected in hidden layer space (randomly generated its neural network version) for models without direct links along with original input data links. The targets encoded multi-class problem according to 1-of-J encoding (J number classes), which implies...
This paper explores the boosting ridge (BR) framework in extreme learning machine (ELM) community and presents a novel model that trains base learners as global ensemble. In context of Extreme Learning Machine single-hidden-layer networks, nodes hidden layer are preconfigured before training, optimisation is performed on weights output layer. The previous implementation BR ensemble with ELM (BRELM) fix for all ELMs. method generates different coefficients by reducing residual error...
Time series clustering is the process of grouping time with respect to their similarity or characteristics. Previous approaches usually combine a specific distance measure for and standard method. However, these do not take different subsequences each into account, which can be used better compare objects dataset. In this paper, we propose novel technique based on two stages. first step, least squares polynomial segmentation procedure applied series, growing window that returns...