arXiv subjects

Rafael A. Moral

Publications and source records attributed to Rafael A. Moral.

13 recordsLinked to original sources

arXiv2025

Bayesian Additive Main Effects and Multiplicative Interaction Models using Tensor Regression for Multi-environmental Trials

We propose a Bayesian tensor regression model to accommodate the effect of multiple factors on phenotype prediction. We adopt a set of prior distributions that resolve identifiability issues that may arise between the parameters in the model. Further, we incorporate a spike-and-slab structure that identifies which interactions are relevant for inclusion in the linear predictor, even when they form a subset of the available variables. Simulation experiments show that our method outperforms previous related models and machine learning algorithms under different sample sizes and degrees of complexity. We further explore the applicability of our model by analysing real-world data related to wheat production across Ireland from 2010 to 2019. Our model performs competitively and overcomes key limitations found in other analogous approaches. Finally, we adapt a set of visualisations for the posterior distribution of the tensor effects that facilitate the identification of optimal interactions between the tensor variables, whilst accounting for the uncertainty in the posterior distribution.

stat.ML↗

arXiv2025

Estimating the number of household TV profiles based in customer behaviour using Gaussian mixture model averaging

TV customers today face many choices from many live channels and on-demand services. Providing a personalised experience that saves customers time when discovering content is essential for TV providers. However, a reliable understanding of their behaviour and preferences is key. When creating personalised recommendations for TV, the biggest challenge is understanding viewing behaviour within households when multiple people are watching. The objective is to detect and combine individual profiles to make better-personalised recommendations for group viewing. Our challenge is that we have little explicit information about who is watching the devices at any time (individuals or groups). Also, we do not have a way to combine more than one individual profile to make better recommendations for group viewing. We propose a novel framework using a Gaussian mixture model averaging to obtain point estimates for the number of household TV profiles and a Bayesian random walk model to introduce uncertainty. We applied our approach using data from real customers whose TV-watching data totalled approximately half a million observations. Our results indicate that combining our framework with the selected features provides a means to estimate the number of household TV profiles and their characteristics, including shifts over time and quantification of uncertainty.

stat.ME↗

arXiv2025

Towards species' classification of the \textit{Anastrepha pseudoparallela} group

Identifying \textit{Anastrepha} species from the \textit{pseudoparallela} group is problematic due to morphological similarities among species and a broad geographic variation. This group comprises $31$ species of fruit flies and pests affecting passion fruit crops. Identifying these species utilises the morphological characteristics of the specimens' wings and aculeus tips. Considering the importance of identifying these species of this group and challenges due to the low data availability and class imbalance, this paper contributes to automating the classification of the species \textit{Anastrepha chiclayae} Greene, \textit{Anastrepha consobrina} (Loew), \textit{Anastrepha curitibana} Araújo, Norrbom \& Savaris, \textit{Anastrepha curitis} Stone, and \textit{Anastrepha pseudoparallela} (Loew) using deep Learning methods and designing new features based on wing structures. We explored transfer learning solutions and proposed a new set of features based on each wing's structure. We used the dual annealing algorithm to optimise the hyperparameters of Deep Neural Networks, Random Forests, Decision Trees, and Support Vector Machines combined with autoencoders and SMOTE algorithms to account for class imbalance. We tested our approach with a dataset of $127$ high-quality images belonging to a total of $5$ species and used three-fold cross validation for training, tuning and testing, encompassing six permutations of those to assess the performance of the learning algorithms. Our findings demonstrate that our novel approach, which combines feature extraction and machine learning techniques, can improve the species classification accuracy for rare \textit{Anastrepha pseudoparallela} group specimens, with the SMOTE and Random Forests algorithms leading to the average performance of $0.72$ in terms of the mean of the individual accuracies considering all species.

q-bio.QM↗

arXiv2024

Accounting for shared covariates in semi-parametric Bayesian additive regression trees

We propose some extensions to semi-parametric models based on Bayesian additive regression trees (BART). In the semi-parametric BART paradigm, the response variable is approximated by a linear predictor and a BART model, where the linear component is responsible for estimating the main effects and BART accounts for non-specified interactions and non-linearities. Previous semi-parametric models based on BART have assumed that the set of covariates in the linear predictor and the BART model are mutually exclusive in an attempt to avoid poor coverage properties and reduce bias in the estimates of the parameters in the linear predictor. The main novelty in our approach lies in the way we change the tree-generation moves in BART to deal with this bias and resolve non-identifiability issues between the parametric and non-parametric components, even when they have covariates in common. This allows us to model complex interactions involving the covariates of primary interest, both among themselves and with those in the BART component. Our novel method is developed with a view to analysing data from an international education assessment, where certain predictors of students' achievements in mathematics are of particular interpretational interest. Through additional simulation studies and another application to a well-known benchmark dataset, we also show competitive performance when compared to regression models, alternative formulations of semi-parametric BART, and other tree-based methods. The implementation of the proposed method is available at \url{https://github.com/ebprado/CSP-BART}.

stat.ML↗

arXiv2024

Understanding Social Immunity in Ants: A Markovian Approach to Collective Cleaning Strategies

Understanding social immunity mechanisms in ant colonies remains crucial to comprehending the evolution of defense strategies in eusocial organisms. This study assumes the absence of the role of memory in the ants' defense strategy, considering that they can make a new exploration of collective cleaning behaviors. We investigate how worker interactions and previous behaviors influence the evolution of social immunity strategies in response to the presence of pathogens and vulnerable colony members. In this context, with the application of Markov transition models it was possible to describe changes in cleaning behavior over time influenced by the presence of vulnerable members and treatment with an entomopathogenic fungus. We found a significant effect on ant behavior when exposed to \it{Metarhizium anisopliae} when a fungus garden fragment and one larva were present. Our findings confirm a link between prophylactic cleaning and the presence of vulnerable members. Remarkably, distinct behaviors and transition times vary among treatments, revealing workers' adaptability to threats. Allogrooming shows adaptive changes when exposed to pathogens, potentially affecting pathogen transmission. In addition, our study elucidates the intricate interaction between internal and external factors shaping worker behavior. The influence of environmental context on decision-making principles emerges, emphasizing the importance of both intrinsic colony organization and external threats. By using a Markov model to understand ant hygiene behavior, we offer insights into social immunity mechanisms in eusocial organisms. Deciphering collective cleaning strategies can aid in understanding disease dynamics and decision-making processes in complex societies.

q-bio.QM↗

arXiv2023

Forecasting insect abundance using time series embedding and machine learning

Implementing insect monitoring systems provides an excellent opportunity to create accurate interventions for insect control. However, selecting the appropriate time for an intervention is still an open question due to the inherent difficulty of implementing on-site monitoring in real-time. This decision is even more critical with insect species that can abruptly increase population size. A possible solution to enhance decision-making is to apply forecasting methods to predict insect abundance. However, another layer of complexity is added when other covariates are considered in the forecasting, such as climate time series collected along the monitoring system. Multiple possible combinations of climate time series and their lags can be used to build a forecasting method. Therefore, this research paper proposes a new approach to address this problem by combining statistics, machine learning, and time series embedding. We used two datasets containing a time series of aphids and climate data collected weekly in Coxilha and Passo Fundo municipalities in Southern Brazil for eight years. We conduct a simulation study based on a probabilistic autoregressive model with exogenous time series based on Poisson and negative binomial distributions to check the influence of incorporating climate time series on the performance of our approach. We pre-processed the data using our newly proposed approach and more straightforward approaches commonly used to train machine learning algorithms in time series problems. We evaluate the performance of the selected machine algorithms by looking at the Root Mean Squared Error obtained using one-step-ahead forecasting. Based on Random Forests, Lasso-regularised linear regression, and LightGBM regression algorithms, our novel approach yields competitive forecasts while automatically selecting insect abundances, climate time series and their lags to aid forecasting.

q-bio.QM↗

arXiv2023

Going beyond richness: Modelling the BEF relationship using species identity, evenness, richness and species interactions via the DImodels R package

BEF studies aim to understand how ecosystems respond to a gradient of species diversity. Diversity-Interactions (DI) models are suitable for analysing the BEF relationship. These models relate an ecosystem function response of a community to the identity of the species in the community, their evenness (proportions) and interactions. The number of species in the community (richness) is also implicitly modelled through this approach. It is common in BEF studies to model an ecosystem function as a function of richness; while this can uncover trends in the BEF relationship, by definition, species diversity is much broader than richness alone, and important patterns in the BEF relationship may remain hidden. In this paper, we introduce the DImodels R package for implementing DI models. We also compare DI models to traditional modelling approaches to highlight the advantages of using a multi-dimensional definition of species diversity. We show that using DI models can lead to considerably improved model fit over other methods; it does this by incorporating variation due to the multiple facets of species diversity. Predicting from a DI model is not limited to the study design points, the model can extrapolate to predict for any species composition and proportions (assuming there is sufficient coverage of this space in the study design). Expressing the BEF relationship as a function of richness alone can be useful to capture overall trends. However, collapsing the multiple dimensions of species diversity to a single dimension (such as richness) can result in valuable ecological information being lost. DI modelling provides a framework to test the multiple components of species diversity in the BEF relationship. It facilitates uncovering a deeper ecological understanding of the BEF relationship and can lead to enhanced inference.

q-bio.PE↗

arXiv2022

Pattern-Based Prediction of Population Outbreaks

Insect outbreaks are biotic disturbances in forests and agroecosystems that cause economic and ecological damage. This phenomenon depends on a variety of biological and physical factors. The complexity and practical importance of the issue have made the problem of predicting outbreaks a focus of recent research. Here, we propose the Pattern-Based Prediction (PBP) method for predicting population outbreaks. It is based on the Alert Zone Procedure, combined with elements from machine learning. It uses information on previous time series values that precede an outbreak event as predictors of future outbreaks, which can be useful when monitoring pest species. We illustrate the methodology using simulated datasets and real time series data obtained by monitoring aphids in wheat crops in Southern Brazil. We obtained an average test accuracy of $84.6\%$ in the simulation studies implemented with stochastic models, and $95.0\%$ for predicting outbreaks using the real dataset. This shows the feasibility of the PBP method in predicting outbreaks in population dynamics. We benchmarked our results against established state-of-the-art machine learning methods, namely Support Vector Machines, Deep Neural Networks, Long Short Term Memory and Random Forests. The PBP method yielded a competitive performance, associated with higher true-positive rates in most comparisons, while being able to provide interpretability rather than being a black-box method. This is an improvement over current state-of-the-art machine learning tools, especially when being used by non-specialists, such as ecologists aiming to use a quantitative approach for pest monitoring. We provide open-source code to implement the PBP method in Python, through the \texttt{pypbp} package, which may be directly downloaded from the Python Package Index server or accessed through \url{https://pypbp-documentation.readthedocs.io}

q-bio.QM↗

arXiv2022

Profiling Television Watching Behaviour Using Bayesian Hierarchical Joint Models for Time-to-Event and Count Data

Customer churn prediction is a valuable task in many industries. In telecommunications it presents great challenges, given the high dimensionality of the data, and how difficult it is to identify underlying frustration signatures, which may represent an important driver regarding future churn behaviour. Here, we propose a novel Bayesian hierarchical joint model that is able to characterise customer profiles based on how many events take place within different television watching journeys, and how long it takes between events. The model drastically reduces the dimensionality of the data from thousands of observations per customer to 11 customer-level parameter estimates and random effects. We test our methodology using data from 40 BT customers (20 active and 20 who eventually cancelled their subscription) whose TV watching behaviours were recorded from October to December 2019, totalling approximately half a million observations. Employing different machine learning techniques using the parameter estimates and random effects from the Bayesian hierarchical model as features yielded up to 92\% accuracy predicting churn, associated with 100\% true positive rates and false positive rates as low as 14\% on a validation set. Our proposed methodology represents an efficient way of reducing the dimensionality of the data, while at the same time maintaining high descriptive and predictive capabilities. We provide code to implement the Bayesian model at https://github.com/rafamoral/profiling_tv_watching_behaviour.

stat.ME↗

arXiv2022

Variational Inference for Additive Main and Multiplicative Interaction Effects Models

In plant breeding the presence of a genotype by environment (GxE) interaction has a strong impact on cultivation decision making and the introduction of new crop cultivars. The combination of linear and bilinear terms has been shown to be very useful in modelling this type of data. A widely-used approach to identify GxE is the Additive Main Effects and Multiplicative Interaction Effects (AMMI) model. However, as data frequently can be high-dimensional, Markov chain Monte Carlo (MCMC) approaches can be computationally infeasible. In this article, we consider a variational inference approach for such a model. We derive variational approximations for estimating the parameters and we compare the approximations to MCMC using both simulated and real data. The new inferential framework we propose is on average two times faster whilst maintaining the same predictive performance as MCMC.

stat.ML↗

arXiv2021

Modelling excess zeros in count data: A new perspective on modelling approaches

We consider the analysis of count data in which the observed frequency of zero counts is unusually large, typically with respect to the Poisson distribution. We focus on two alternative modelling approaches: Over-Dispersion (OD) models, and Zero-Inflation (ZI) models, both of which can be seen as generalisations of the Poisson distribution; we refer to these as Implicit and Explicit ZI models, respectively. Although sometimes seen as competing approaches, they can be complementary; OD is a consequence of ZI modelling, and ZI is a by-product of OD modelling. The central objective in such analyses is often concerned with inference on the effect of covariates on the mean, in light of the apparent excess of zeros in the counts. Typically the modelling of the excess zeros per se is a secondary objective and there are choices to be made between, and within, the OD and ZI approaches. The contribution of this paper is primarily conceptual. We contrast, descriptively, the impact on zeros of the two approaches. We further offer a novel descriptive characterisation of alternative ZI models, including the classic hurdle and mixture models, by providing a unifying theoretical framework for their comparison. This in turn leads to a novel and technically simpler ZI model. We develop the underlying theory for univariate counts and touch on its implication for multivariate count data.

stat.ME↗

arXiv2021

Bayesian Additive Regression Trees with Model Trees

Bayesian Additive Regression Trees (BART) is a tree-based machine learning method that has been successfully applied to regression and classification problems. BART assumes regularisation priors on a set of trees that work as weak learners and is very flexible for predicting in the presence of non-linearity and high-order interactions. In this paper, we introduce an extension of BART, called Model Trees BART (MOTR-BART), that considers piecewise linear functions at node levels instead of piecewise constants. In MOTR-BART, rather than having a unique value at node level for the prediction, a linear predictor is estimated considering the covariates that have been used as the split variables in the corresponding tree. In our approach, local linearities are captured more efficiently and fewer trees are required to achieve equal or better performance than BART. Via simulation studies and real data applications, we compare MOTR-BART to its main competitors. R code for MOTR-BART implementation is available at https://github.com/ebprado/MOTR-BART.

stat.ML↗

arXiv2018

Modeling data with zero inflation and overdispersion using GAMLSSs

Count data with high frequencies of zeros are found in many areas, specially in biology. Statistical models to analyze such data started to be developed in the 80s and are still a topic of active research. Such models usually assume a response distribution that belongs to the exponential family of distributions and the analysis is performed under the generalized linear models framework. However, the generalized additive models for location, scale and shape (GAMLSSs) represent a more general class of univariate models that can also be used to model zero inflated data. In this paper, the analysis of a data set with excess of zeros and overdispersion is described using GAMLSSs. Specific GAMLSSs' tools were used in the analysis, which enhanced model comparison and eased the interpretation of results.

stat.AP↗