Search

Background Human leptospirosis is responsible for great losses and deaths, especially in developing countries, which can be mitigated by knowing the correct health indicators and climate influence on the disease. Methods Leptospirosis cases and deaths, population and precipitation were recovered from different databases (2007–2019). Annual incidence, mortality and case fatality rates (CFRs) of human leptospirosis and average precipitation were calculated for Brazil and its regions. Time series analysis using an moving average with external variable (ARMAX) model was used to analyse the monthly contribution and precipitation influence over leptospirosis cases for each Brazilian region and for the whole country. A forecast model to predict cases for 2020 was created for Brazil. Results Human leptospirosis exhibited heterogeneous distribution among Brazilian regions, with most cases occurring during the rainy season and precipitation influenced the disease occurrence in all regions but the South. The forecast model predicted 3276.99 cases for 2020 (mean absolute percentage error 14.680 and root mean square error 53.013). Considering the annual average for the period, the leptospirosis incidence was 1913 cases per 100 000 inhabitants, mortality was 0.168 deaths per 100 000 inhabitants and the CFR was 8.83%. Conclusions The models built can be useful for planning leptospirosis surveillance and control actions for the whole country and its regions and, together with the health indicators, revealed no uniform epidemiological situation of leptospirosis in Brazil.

Forecasting models in the financial market generally use quantitative time-series data. However, external factors can influence data in time-series, such as weather events, economic crises, and the foreign exchange market. This information is not explicit in the time-series and can influence the prediction of the variable values. Textual data can be a source of knowledge about external factors and is potentially helpful for time-series forecasting models. Some studies have presented text mining techniques to combine textual and time-series data. However, the existing representations have limitations, such as the curse of dimensionality and sparse data. This work investigates the finite use of domain-specific terms to investigate these problems by representing textual data with low dimensional space. We consider thirty-three keywords that are potentially important in the domain to enrich time-series using text mining techniques. Four regression models were applied to the representation proposed to predict the future daily price of corn and soybeans. The experimental setup considers a real market scenario, in which the daily sliding window strategy and step-forward forecast were used. The representation proposed has better accuracy in some forecasting scenarios. The results indicate that text data are a promising alternative for enriching time-series representations and reducing uncertainty forecasting models.•We show an approach to enriching time-series using domain-specific terms;•Representation proposed combines quantitative data with qualitative market factors;•Regression Models to learn a forecasting function from enriched time-series.

In this paper, we introduce the concept of learning to sense, which aims to emulate a complex characteristic of human reasoning: the ability to monitor and understand a set of interdependent events for decision-making processes. Event datasets are composed of textual data and spatio-temporal features that determine where and when a given phenomenon occurred. In learning to sense, related events are mapped closely to each other in a semantic vector space, thereby identifying that they contain similar contextual meaning. However, learning a semantic vector space that satisfies both textual similarities and spatio-temporal constraints is a crucial challenge for event analysis and sensing. This paper investigates a Semantic Variational Autoencoder (SVAE) to fine-tune pre-trained embeddings according to both textual and spatio-temporal events of the class of interest. Experiments involving more than one hundred sensors show that our SVAE outperforms a competitive one-class classification baseline. Moreover, our proposal provides desirable learning requirements to sense scenarios, such as visualization of the sensor decision function and heat maps with the sensor’s geographic impact.

Combination methods have performed well in time series forecast competitions. This study proposes a simple but general methodology for combining time series forecast methods. Weights are calculated using a cross-validation scheme that assigns greater weights to methods with more accurate in-sample predictions. The methodology was used to combine forecasts from the Theta, exponential smoothing, and ARIMA models, and placed fifth in the M4 Competition for both point and interval forecasting.

Background In Brazil, a substantial number of coronavirus disease (COVID-19) cases and deaths have been reported. It has become the second most affected country worldwide, as of June 9, 2020. Official Brazilian government sources present contradictory data on the impact of the disease; thus, it is possible that the actual number of infected individuals and deaths in Brazil is far larger than those officially reported. It is very likely that the actual spread of the disease has been underestimated. Objective This study investigates the underreporting of cases and deaths related to COVID-19 in the most affected cities in Brazil, based on public data available from official Brazilian government internet portals, to identify the actual impact of the pandemic. Methods We used data from historical deaths due to respiratory problems and other natural causes from two public portals: DATASUS (Department of Informatics of the Unified Healthcare System) (2010-2018) and the Brazilian Transparency Portal of Civil Registry (2019-2020). These data were used to build time-series models (modular regressions) to predict the expected mortality patterns for 2020. The forecasts were used to estimate the possible number of deaths that were incorrectly registered during the pandemic and posted on government internet portals in the most affected cities in the country. Results Our model found a significant difference between the real and expected values. The number of deaths due to severe acute respiratory syndrome (SARS) was considerably higher in all cities, with increases between 493% and 5820%. This sudden increase may be associated with errors in reporting. An average underreporting of 40.68% (range 25.9%-62.7%) is estimated for COVID-19–related deaths. Conclusions The significant rates of underreporting of deaths analyzed in our study demonstrate that officially released numbers are much lower than actual numbers, making it impossible for the authorities to implement a more effective pandemic response. Based on analyses carried out using different fatality rates, it can be inferred that Brazil’s epidemic is worsening, and the actual number of infectees could already be between 1 to 5.4 million.

Summary This paper reviews and extends several aspects of the analysis of linear combinations of time series. Special cases are temporal and contemporaneous aggregations and systematic sampling. We present some simple examples, a unified notation, references to the literature, and some general results for linear combinations of scalar and vector time series. For basic time series following ARIMA models in scalar cases we derive the ARIMA models of the linear combinations as functions of those of the basic series in both the nonseasonal and seasonal cases. For vector time series we compare the forecast efficiencies of two alternative approaches: first model and forecast and then form the linear combination, and first form the linear combination and then model and forecast; for this analysis we use the moving average representation of a stationary time series. A final section contains an application to monthly data on milk production and milk productivity series for the State of Siio Paulo, Brazil.