Yuzhi Cai

Love Thy Neighbour: Automatic Animal Behavioural Classification of Acceleration Data Using the K-Nearest Neighbour Algorithm

Researchers hoping to elucidate the behaviour of species that aren’t readily observed are able to do so using biotelemetry methods. Accelerometers in particular are proving particularly effective and have been used on terrestrial, aquatic and volant species with success. In the past, behavioural modes were detected in accelerometer data through manual inspection, but with developments in technology, modern accelerometers now record at frequencies that make this impractical. In light of this, some researchers have suggested the use of various machine learning approaches as a means to classify accelerometer data automatically. We feel uptake of this approach by the scientific community is inhibited for two reasons; 1) Most machine learning algorithms require selection of summary statistics which obscure the decision mechanisms by which classifications are arrived, and 2) they are difficult to implement without appreciable computational skill. We present a method which allows researchers to classify accelerometer data into behavioural classes automatically using a primitive machine learning algorithm, k-nearest neighbour (KNN). Raw acceleration data may be used in KNN without selection of summary statistics, and it is easily implemented using the freeware program R. The method is evaluated by detecting 5 behavioural modes in 8 species, with examples of quadrupedal, bipedal and volant species. Accuracy and Precision were found to be comparable with other, more complex methods. In order to assist in the application of this method, the script required to run KNN analysis in R is provided. We envisage that the KNN method may be coupled with methods for investigating animal position, such as GPS telemetry or dead-reckoning, in order to implement an integrated approach to movement ecology research.

Bayesian inference for ion–channel gating mechanisms directly from single–channel recordings, using Markov chain Monte Carlo

The gating mechanism of a single–ion channel is usually modelled by a finite–state–space continuous–time Markov chain. The patch–clamp technique enables the experimenter to record the current flowing across a single–ion channel. In practice, the current is corrupted by noise and low–pass filtering, and is sampled with a typically very short sampling interval. We present a method for performing Bayesian inference about parameters governing the underlying single–channel gating mechanism and the recording process, directly from such single–channel recordings. Our procedure uses a technique known as Markov chain Monte Carlo, which involves constructing a Markov chain whose equilibrium distribution is given by the posterior distribution of the unknown parameters given the observed data. Simulation of that Markov chain then enables the investigator to estimate the required posterior distribution. As well as providing a method of estimating the transition rates of the underlying Markov chain used to model the single–channel gating mechanism and the means and variances of open and closed conductance levels, the output from our Markov chain Monte Carlo simulations can also be used to estimate single–channel properties, such as the mean lengths of open and closed sojourn times, and to reconstruct the unobserved quantal signal which indicates whether the channel is open or closed. The theory is illustrated by several numerical examples taken mainly from the ion–channel literature.

Bayesian nonparametric quantile regression using splines

A new technique based on Bayesian quantile regression that models the dependence of a quantile of one variable on the values of another using a natural cubic spline is presented. Inference is based on the posterior density of the spline and an associated smoothing parameter and is performed by means of a Markov chain Monte Carlo algorithm. Examples of the application of the new technique to two real environmental data sets and to simulated data for which polynomial modelling is inappropriate are given. An aid for making a good choice of proposal density in the Metropolis-Hastings algorithm is discussed. The new nonparametric methodology provides more flexible modelling than the currently used Bayesian parametric quantile regression approach.

Quantile self-exciting threshold autoregressive time series models

In this paper we present a Bayesian approach to quantile self-exciting threshold autoregressive time series models. The simulation work shows that the method can deal very well with nonstationary time series with very large, but not necessarily symmetric, variations. The methodology has also been applied to the growth rate of US real GNP data and some interesting results have been obtained. Copyright 2007 The Author Journal compilation 2007 Blackwell Publishing Ltd.

Wild state secrets: ultra-sensitive measurement of micro-movement can reveal internal processes in animals

Assessment of animal internal “state” – which includes hormonal, disease, nutritional, and emotional states – is normally considered the province of laboratory work, since its determination in animals in the wild is considered more difficult. However, we show that accelerometers attached externally to animals as diverse as elephants, cockroaches, and humans display consistent signal differences in micro-movement that are indicative of internal state. Originally used to elucidate the behavior of wild animals, accelerometers also have great potential for highlighting animal actions, which are considered as responses stemming from the interplay between internal state and external environment. Advances in accelerometry may help wildlife managers understand how internal state is linked to behavior and movement, and thus clarify issues ranging from how animals cope with the presence of newly constructed roads to how diseased animals might change movement patterns and therefore modulate disease spread.

Perfect simulation for correlated Poisson random variables conditioned to be positive

In this paper we present a perfect simulation method for obtaining perfect samples from collections of correlated Poisson random variables conditioned to be positive. We show how to use this method to produce a perfect sample from a Boolean model conditioned to cover a set of points: in W.S. Kendall and E. Thönnes (Pattern Recognition 32(9): 1569–1586, 1999), this special case was treated in a more complicated way. The method is applied to several simple examples where exact calculations can be made, so as to check correctness of the program using χ2-tests, and some small-scale experiments are carried out to explore the behaviour of the conditioned Boolean model.

A new Bayesian approach to quantile autoregressive time series model estimation and forecasting

This paper proposes a Bayesian approach to quantile autoregressive (QAR) time series model estimation and forecasting. We establish that the joint posterior distribution of the model parameters and future values is well defined. The associated Markov chain Monte Carlo algorithm for parameter estimation and forecasting converges to the posterior distribution quickly. We also present a combining forecasts technique to produce more accurate out-of-sample forecasts by using a weighted sequence of fitted QAR models. A moving window method to check the quality of the estimated conditional quantiles is developed. We verify our methodology using simulation studies and then apply it to currency exchange rate data. The results obtained show that an unequally weighted combining method performs better than other forecasting methodology.

Autoregression with Non-Gaussian Innovations

Many economics and finance time series are non-Gaussian. In this paper, we propose a Bayesian approach to non-Gaussian autoregressive time series models via quantile functions. This approach is parametric, so we also compare the proposed parametric approach with a semi-parametric approach. Simulation studies and applications to real time series show that this method works very well.

Multi‐variate time‐series simulation

In this article we present a method for simulating a multi-variate time series via a vector auto regressive moving average (p, q) process. We also carried out two simulation studies to check the performance of the method and applied the methodology to a real sea condition time series. All results show that the method works very well in practice.

Modelling beach-structure interaction using a Heaviside technique: application and validation

ABSTRACT Valsamidis, A., Cai, Y., Reeve, D.E., 2013. Modelling beach-structure interaction using a Heaviside technique: application and validation In this study, an analytical solution, based on a Heaviside technique, is developed to model the shoreline evolution in the vicinity of a groyne due to a random sequence of waves. The beach at Borth, Wales, UK was used as a case-study. A wave time-series covering a time period of about 12 years, was used to test the performance of a recently constructed coastal defence scheme. Transformations of the wave time-series from offshore to nearshore were performed using a semi-empirical procedure. Three different wave breaking formulae were independently applied to the wave model, and their effects to the consequent shoreline evolution were investigated. In addition, three different longshore transport formulae were compared. These were the CERC, the Kamphuis and the Bayram formulae. Results showed that the CERC formula predicted a significantly greater amount of sediment transport and hence erosion on the downdrift side of the groyne while the models based on Kamphuis and the Bayram formulae gave comparable results. All the results exhibited a strong sensitivity to the temporal resolution of the forcing. Finally, some sensitivity to the treatment of wave breaking was found.