We propose a novel data-driven machine learning method using long short-term memory (LSTM)-based multi-stage forecasting for influenza forecasting. The novel aspects of the method include the following: 1) the introduction of LSTM method to capture the temporal dynamics of seasonal flu and 2) a technique to capture the influence of external variables that includes the geographical proximity and climatic variables such as humidity, temperature, precipitation, and sun exposure. The proposed model is compared against two state-of-the-art techniques using two publicly available datasets. Our proposed method performs better than the existing well-known influenza forecasting methods. The results offer a promising direction in terms of both using the data-driven forecasting methods and capturing the influence of spatio-temporal and environmental factors to improve influenza forecasting.INDEX TERMS Influenza forecasting, LSTM, recurrent neural networks, spatio-temporal data, time series forecasting.

Low frequency sound has increased in the Northeast Pacific Ocean over the past 60 yr [Ross (1993) Acoust. Bull. 18, 5–8; (2005) IEEE J. Ocean. Eng. 30, 257–261; Andrew, Howe, Mercer, and Dzieciuch (2002) J. Acoust. Soc. Am. 129, 642–651; McDonald, Hildebrand, and Wiggins (2006) J. Acoust. Soc. Am. 120, 711–717; Chapman and Price (2011) J. Acoust. Soc. Am. 129, EL161–EL165] and in the Indian Ocean over the past decade, [Miksis-Olds, Bradley, and Niu (2013) J. Acoust. Soc. Am. 134, 3464–3475]. More recently, Andrew, Howe, and Mercer's [(2011) J. Acoust. Soc. Am. 129, 642–651] observations in the Northeast Pacific show a level or slightly decreasing trend in low frequency noise. It remains unclear what the low frequency trends are in other regions of the world. In this work, data from the Comprehensive Nuclear-Test Ban Treaty Organization International Monitoring System was used to examine the rate and magnitude of change in low frequency sound (5–115 Hz) over the past decade in the South Atlantic and Equatorial Pacific Oceans. The dominant source observed in the South Atlantic was seismic air gun signals, while shipping and biologic sources contributed more to the acoustic environment at the Equatorial Pacific location. Sound levels over the past 5–6 yr in the Equatorial Pacific have decreased. Decreases were also observed in the ambient sound floor in the South Atlantic Ocean. Based on these observations, it does not appear that low frequency sound levels are increasing globally.

Abstract-We provide data-driven machine learning methods that are capable of making real-time influenza forecasts that integrate the impacts of climatic factors and geographical proximity to achieve better forecasting performance. The key contributions of our approach are both applying deep learning methods and incorporation of environmental and spatio-temporal factors to improve the performance of the influenza forecasting models. We evaluate the method on Influenza Like Illness (ILI) counts and climatic data, both publicly available data sets. Our proposed method outperforms existing known influenza forecasting methods in terms of their Mean Absolute Percentage Error and Root Mean Square Error. The key advantages of the proposed data-driven methods are as following: (1) The deep-learning model was able to effectively capture the temporal dynamics of flu spread in different geographical regions, (2) The extensions to the deep-learning model capture the influence of external variables that include the geographical proximity and climatic variables such as humidity, temperature, precipitation and sun exposure in future stages, (3) The model consistently performs well for both the city scale and the regional scale on the Google Flu Trends (GFT) and Center for Disease Control (CDC) flu counts. The results offer a promising direction in terms of both datadriven forecasting methods and capturing the influence of spatio-temporal and environmental factors for influenza forecasting methods.

No abstract

Oceanic ambient noise is a dynamic mixture of biologic, geophysical, and anthropogenic sound sources. A goal of research is to put some order in this cacophony of information, understand the received spectral content and determine the primary contributors to the ambient noise. This paper compares three methods to assist in that process (with emphasis on noise correlation techniques): noise correlation matrices, manual selection of noise spectra, and principal component analysis. Comparison followed a common process: selection of a replica set (best termed a characteristic subset of noise spectra), which are used to recreate the original noise field for comparison and consequent decision as to whether that replica set represented the noise measurements adequately. Conclusions of this study are (1) noise correlation matrices provide the best definition of the spectra that represent a particular source and offer potential in organizing and identifying specific noise source content. (2) Manual sorting of noise spectra, while able to identify specific events easily, is both labor intensive, given the quantity of data available; and suffers from incorrect interpretation of multiple competing sound sources, when present. (3) Principal component analysis provides the best reconstruction of measured noise, but has difficulty linking components to physical source mechanisms.