Time Series Segmentation Based on Stationarity Analysis to Improve New Samples Prediction.

A wide range of applications based on sequential data, named time series, have become increasingly popular in recent years, mainly those based on the Internet of Things (IoT). Several different machine learning algorithms exploit the patterns extracted from sequential data to support multiple tasks. However, this data can suffer from unreliable readings that can lead to low accuracy models due to the low-quality training sets available. Detecting the change point between high representative segments is an important ally to find and thread biased subsequences. By constructing a framework based on the Augmented Dickey-Fuller (ADF) test for data stationarity, two proposals to automatically segment subsequences in a time series were developed. The former proposal, called Change Detector segmentation, relies on change detection methods of data stream mining. The latter, called ADF-based segmentation, is constructed on a new change detector derived from the ADF test only. Experiments over real-file IoT databases and benchmarks showed the improvement provided by our proposals for prediction tasks with traditional Autoregressive integrated moving average (ARIMA) and Deep Learning (Long short-term memory and Temporal Convolutional Networks) methods. Results obtained by the Long short-term memory predictive model reduced the relative prediction error from 1 to 0.67, compared to time series without segmentation.

Quantifying the impact of non-stationarity in reinforcement learning-based traffic signal control.

In reinforcement learning (RL), dealing with non-stationarity is a challenging issue. However, some domains such as traffic optimization are inherently non-stationary. Causes for and effects of this are manifold. In particular, when dealing with traffic signal controls, addressing non-stationarity is key since traffic conditions change over time and as a function of traffic control decisions taken in other parts of a network. In this paper we analyze the effects that different sources of non-stationarity have in a network of traffic signals, in which each signal is modeled as a learning agent. More precisely, we study both the effects of changing the context in which an agent learns (e.g., a change in flow rates experienced by it), as well as the effects of reducing agent observability of the true environment state. Partial observability may cause distinct states (in which distinct actions are optimal) to be seen as the same by the traffic signal agents. This, in turn, may lead to sub-optimal performance. We show that the lack of suitable sensors to provide a representative observation of the real state seems to affect the performance more drastically than the changes to the underlying traffic patterns.

A generalized transition model for grouped longitudinal categorical data.

Transition models are an important framework that can be used to model longitudinal categorical data. They are particularly useful when the primary interest is in prediction. The available methods for this class of models are suitable for the cases in which responses are recorded individually over time. However, in many areas, it is common for categorical data to be recorded as groups, that is, different categories with a number of individuals in each. As motivation we consider a study in insect movement and another in pig behaviou. The first study was developed to understand the movement patterns of female adults of Diaphorina citri, a pest of citrus plantations. The second study investigated how hogs behaved under the influence of environmental enrichment. In both studies, the number of individuals in different response categories was observed over time. We propose a new framework for considering the time dependence in the linear predictor of a generalized logit transition model using a quantitative response, corresponding to the number of individuals in each category. We use maximum likelihood estimation and present the results of the fitted models under stationarity and non-stationarity assumptions, and use recently proposed tests to assess non-stationarity. We evaluated the performance of the proposed model using simulation studies under different scenarios, and concluded that our modeling framework represents a flexible alternative to analyze grouped longitudinal categorical data.

The Color of Noise and Weak Stationarity at the NREM to REM Sleep Transition in Mild Cognitive Impaired Subjects.

In Older Adults (OAs), Electroencephalogram (EEG) slowing in frontal lobes and a diminished muscle atonia during Rapid Eye Movement sleep (REM) have each been effective tracers of Mild Cognitive Impairment (MCI), but this relationship remains to be explored by non-linear analysis. Likewise, data provided by EEG, EMG (Electromyogram) and EOG (Electrooculogram)-the three required sleep indicators-during the transition from REM to Non-REM (NREM) sleep have not been related jointly to MCI. Therefore, the main aim of the study was to explore, with results for Detrended Fluctuation Analysis (DFA) and multichannel DFA (mDFA), the Color of Noise (CN) at the NREM to REM transition in OAs with MCI vs. subjects with good performances. The comparisons for the transition from NREM to REM were made for each group at each cerebral area, taking bilateral derivations to evaluate interhemispheric coupling and anteroposterior and posterior networks. In addition, stationarity analysis was carried out to explore if the three markers distinguished between the groups. Neuropsi and the Mini-Mental State Examination (MMSE) were administered, as well as other geriatric tests. One night polysomnography was applied to 6 OAs with MCI (68.1 ± 3) and to 7 subjects without it (CTRL) (64.5 ± 9), and pre-REM and REM epochs were analyzed for each subject. Lower scores for attention, memory and executive funcions and a greater index of arousals during sleep were found for the MCI group. Results confirmed that EOGs constituted significant markers of MCI, increasing the CN for the MCI group in REM sleep. The CN of the EEG from the pre-REM to REM was higher for the MCI group vs. the opposite for the CTRL group at frontotemporal areas. Frontopolar interhemispheric scaling values also followed this trend as well as right anteroposterior networks. EMG Hurst values for both groups were lower than those for EEG and EOG. Stationarity analyses showed differences between stages in frontal areas and right and left EOGs for both groups. These results may demonstrate the breakdown of fractality of areas especially involved in executive functioning and the way weak stationarity analyses may help to distinguish between sleep stages in OAs.

Current and historical climate signatures to deconstructed tree species richness pattern in South America / Efeitos dos climas atual e histórico no padrão desconstruído de riqueza de espécies arbóreas na região Neotropical

The purpose of this study was to investigate the importance of present and historical climate as determinants of current species richness pattern of forestry trees in South America. The study predicted the distribution of 217 tree species using Maxent models, and calculated the potential species richness pattern, which was further deconstructed based on range sizes and modeled against current and historical climates predictors using Geographically Weighted Regressions (GWR) analyses. The current climate explains more of the wide-ranging species richness patterns than that of the narrow-ranging species, while the historical climate explained an equally small amount of variance for both narrow-and-wide ranging tree species richness patterns. The richness deconstruction based on range size revealed that the influences of current and historical climate hypotheses underlying patterns in South American tree species richness differ from those found in the Northern Hemisphere. Notably, the historical climate appears to be an important determinant of richness only in regions with marked climate changes and proved Pleistocenic refuges, while the current climate predicts the species richness across those Neotropical regions, with non-evident refuges in the Last Glacial Maximum. Thus, this study's analyses show that these climate hypotheses are complementary to explain the South American tree species richness.

O objetivo deste estudo foi testar qual dos climas, atual ou histórico, é o principal preditor do padrão atual de riqueza de espécies arbóreas de interesse comercial. Nós modelamos a distribuição de 217 espécies usando Maxent e usamos esses mapas preditivos para obter o padrão de riqueza de espécies. A riqueza foi desconstruída em relação ao tamanho da distribuição geográfica das espécies e modelada contra os climas atual e histórico utilizando Regressões Geograficamente Ponderadas. O clima atual explicou melhor o padrão de riqueza das espécies com ampla distribuição geográfica do que de espécies com distribuição restrita, enquanto o clima histórico explicou a mesma variância na riqueza dos dois grupos de espécies. Nossas análises com plantas sul americanas mostram diferentes relações da riqueza de espécies ampla e restritamente distribuídas com os climas atual e histórico, quando comparado aos resultados encontrados no hemisfério norte. O clima histórico se mostra como importante preditor da riqueza somente em regiões com mudanças climáticas acentuadas e onde ocorreram refúgios Pleistocênicos, enquanto o clima atual é o melhor da riqueza nas regiões Neotropicais sem evidências de refúgios durante o máximo da ultima glaciação. Dessa maneira, nossos resultados indicam que essas hipóteses são complementares para explicar a riqueza predita de espécies arbóreas da América do Sul.

Current and historical climate signatures to deconstructed tree species richness pattern in South America / Efeitos dos climas atual e histórico no padrão desconstruído de riqueza de espécies arbóreas na região Neotropical

The purpose of this study was to investigate the importance of present and historical climate as determinants of current species richness pattern of forestry trees in South America. The study predicted the distribution of 217 tree species using Maxent models, and calculated the potential species richness pattern, which was further deconstructed based on range sizes and modeled against current and historical climates predictors using Geographically Weighted Regressions (GWR) analyses. The current climate explains more of the wide-ranging species richness patterns than that of the narrow-ranging species, while the historical climate explained an equally small amount of variance for both narrow-and-wide ranging tree species richness patterns. The richness deconstruction based on range size revealed that the influences of current and historical climate hypotheses underlying patterns in South American tree species richness differ from those found in the Northern Hemisphere. Notably, the historical climate appears to be an important determinant of richness only in regions with marked climate changes and proved Pleistocenic refuges, while the current climate predicts the species richness across those Neotropical regions, with nonevident refuges in the Last Glacial Maximum. Thus, this study"s analyses show that these climate hypotheses are complementary to explain the South American tree species richness.(AU)

O objetivo deste estudo foi testar qual dos climas, atual ou histَrico, é o principal preditor do padrمo atual de riqueza de espécies arbَreas de interesse comercial. Nَs modelamos a distribuiçمo de 217 espécies usando Maxent e usamos esses mapas preditivos para obter o padrمo de riqueza de espécies. A riqueza foi desconstruيda em relaçمo ao tamanho da distribuiçمo geogrلfica das espécies e modelada contra os climas atual e histَrico utilizando Regressُes Geograficamente Ponderadas. O clima atual explicou melhor o padrمo de riqueza das espécies com ampla distribuiçمo geogrلfica do que de espécies com distribuiçمo restrita, enquanto o clima histَrico explicou a mesma variância na riqueza dos dois grupos de espécies. Nossas anلlises com plantas sul americanas mostram diferentes relaçُes da riqueza de espécies ampla e restritamente distribuيdas com os climas atual e histَrico, quando comparado aos resultados encontrados no hemisfério norte. O clima histَrico se mostra como importante preditor da riqueza somente em regiُes com mudanças climلticas acentuadas e onde ocorreram refْgios Pleistocênicos, enquanto o clima atual é o melhor da riqueza nas regiُes Neotropicais sem evidências de refْgios durante o mلximo da ultima glaciaçمo. Dessa maneira, nossos resultados indicam que essas hipَteses sمo complementares para explicar a riqueza predita de espécies arbَreas da América do Sul.(AU)

Diminished heart beat non-stationarities in congestive heart failure.

Studies on heart rate variability (HRV) have become popular and the possibility of diagnosis based on non-invasive techniques compels us to overcome the difficulties originated on the environmental changes that can affect the signal. We perform a non-parametric segmentation which consists of locating the points where the signal can be split into stationary segments. By finding stationary segments we are able to analyze the size of these segments and evaluate how the signal changes from one segment to another, looking at the statistical moments given in each patch, for example, mean and variance. We analyze HRV data for 15 patients with congestive heart failure (CHF; 11 males, 4 females, age 56±11 years), 18 elderly healthy subjects (EH; 11 males, 7 females, age 50±7 years), and 15 young healthy subjects (YH; 11 females, 4 males, age 31±6 years). Our results confirm higher variance for YH, and EH, while CHF displays diminished variance with p-values <0.01, when compared to the healthy groups, presenting higher HRV in healthy subjects. Moreover, it is possible to distinguish between YH and EH with p < 0.05 through the segmentation outcomes. We found high correlations between the results of segmentation and standard measures of HRV analysis and a connection to results of detrended fluctuation analysis (DFA). The segmentation applied to HRV studies detects aging and pathological conditions effects on the non-stationary behavior of the analyzed groups, promising to contribute in complexity analysis and providing risk stratification measures.