EVENTO
A GENERAL PURPOSE MATHEMATICAL/COMPUTATIONAL MODEL FOR NOWCASTING
Tipo de evento: Exame de Qualificação
In recent years, data sensor networks have been deployed in various domains (medical sciences for patient care using biometric sensors, wildfire detection, meteorology for weather forecasting, satellite imaging for earth and space observation, agricultural lands, transportation systems, urban mobility, etc). The sensors, that include physical and social sensors, are distributed across the globe, capturing and continuously producing an enormous amount of information about a number of real world phenomena. The vast ocean of data collected every day by sensors, provides new possibilities for scientific ambit to extract useful knowledge from data and predict future behaviors on different areas. Despite of these vast amount of available data is available, large part of them are untrustworthy and are not validated by quality assurance/quality control (QA/QC) techniques. Even so, it potentially provides a valuable source of high temporal and spatial resolution, real-time data, especially in regions where few observations currently exist, thereby adding value to science, technology and society. In many different cases traditional sources are not widely available in real-time or at the range of spatio-temporal scales required for numerous applications, such as: flood-water and urban drainage management (e.g. Willems et al., 2012; Arnbjerg-Nielsen et al., 2013), urban heat island monitoring (e.g. Tomlinson et al., 2013), planning and decision-making (e.g. Neirotti et al., 2014), precision farming (e.g. Goodchild, 2007), hazard warning systems (e.g. NRC, 2007), road winter maintenance (e.g. Chapman et al., 2014), climate and health risk assess- ments (e.g. Tomlinson et al., 2011), nowcasting (e.g. Ochoa-Rodriguez et al., 2013), model assimilation and evaluation (e.g. Ashie and Kono, 2011), radar and satellite validation (e.g. Binau, 2012), and other societal applications. Taking into account the need of using multiple data sources and on the other hand the problems associated to them, new challenges during its processing and interpretation must be addressed. Some of them are: detecting untrustworthy information, because such information usually indicates critical, unusual, or suspicious activities; building forecasting (nowcasting) models using heterogeneous data sources and offering sufficient flexibility to make predictions, even when some data sources are not available. In that sense, we propose the development of algorithms and techniques to: (i) evaluate the untrustworthuntrustworthinessy of the data sources, both historical and real time data sources. In (Xiao et al., 2015) they propose a tensor factorization technique to verifyied the untrustworthinessy of a data source, this approach could be extended to other kind of sources with different characteristics.The idea is to extend theirre approach from a more general point of view.(ii) on the other hand, in (Shi et al., 2015) they purpose a framework is proposed to build models and make predictions of the future behavior of some systems, based in Deep Learning theory, more specifically Long-Short Term Memory (LSTM, a kind of Recurrent Neural Network). They use tensor too, to represent the data sets. In that case, we can extend their approach in two directions. First we have a problem with the data sets, many times we don't have information related to some data source, or the data source are untrustworthy, then we can't use this information to make predictions. Using the convolution approach proposed by (Shi et al., 2015) in conjunction with the LSTM we can estimate the values of the missing or untrustworthy sources and use them to make predictions.Second, in theirre approach (Shi et al., 2015), they use a static convolutional matrix to project the past state in the present state without taking into account for example physical variables. Suppose we are trying to predict the movement of a rainfall, the projection of the convolutional matrix could be modified according to the wind velocity and direction. BIBLIOGRAFIA: Antenucci, D., Li, E., Liu, S., Zhang, B., Cafarella, M. J., & Ré, C. (2013). Ringtail: a generalized nowcasting system. Proceedings of the VLDB Endowment, 6(12), 1358-1361.Antenucci, D., Cafarella, M. J., Levenstein, M., Ré, C., & Shapiro, M. (2013). Ringtail: Feature Selection For Easier Nowcasting. In WebDB (pp. 49-54).Atefeh, F., & Khreich, W. (2013). A survey of techniques for event detection in twitter. Computational Intelligence.Castillo, C., Mendoza, M., & Poblete, B. (2013). Predicting information credibility in time-sensitive social media. Internet Research, 23(5), 560-588.D'Andrea, E., Ducange, P., Lazzerini, B., & Marcelloni, F. Real-Time Detection of Traffic From Twitter Stream Analysis.Kalampokis, E., Tambouris, E., & Tarabanis, K. (2013). Understanding the predictive power of social media. Internet Research, 23(5), 544-559.Keis, F., & Wiesner, K. (2014, April). Participatory sensing utilized by an advanced meteorological nowcasting system. In Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP), 2014 IEEE Ninth International Conference on (pp. 1-6). IEEE.Keis, F. (2015). WHITEWinter hazards in terminal environment: An automated nowcasting system for Munich Airport. Meteorologische Zeitschrift, 61-82.Lampos, V., & Cristianini, N. (2012). Nowcasting events from the social web with statistical learning. ACM Transactions on Intelligent Systems and Technology (TIST), 3(4), 72.Le, L., Ferrara, E., & Flammini, A. (2015). On predictability of rare events leveraging social media: a machine learning perspective. arXiv preprint arXiv:1502.05886.Lin, L., Ni, M., He, Q., Gao, J., Sadek, A. W., & Director, T. I. T. I. (2015). Modeling the Impacts of Inclement Weather on Freeway Traffic Speed: An Exploratory Study Utilizing Social Media Data. In Transportation Research Board 94th Annual Meeting (No. 15-4749).Mass, C., & Mass, C. F. (2011). Nowcasting: The Next Revolution in Weather Prediction. Bulletin of the American Meteorological Society.Muller, C. L., Chapman, L., Johnston, S., Kidd, C., Illingworth, S., Foody, G., & Leigh, R. R. (2015). Crowdsourcing for climate and atmospheric sciences: current status and future potential. International Journal of Climatology.Nantes, A., Ngoduy, D., Bhaskar, A., Miska, M., & Chung, E. (2015). Real-time traffic state estimation in urban corridors from heterogeneous data. Transportation Research Part C: Emerging Technologies.Ochoa-Rodriguez, S., Rico-Ramirez, M. A., Jewell, S. A., Schellart, A. N. A., Wang, L., Onof, C., & Maksimovic, C. (2013). Improving rainfall nowcasting and urban runoff forecasting through dynamic radar-raingauge rainfall adjustment. In Proceeding of 7th International Conference on Sewer Processes and Networks.Savage, M. J. (2015). Nowcasting daily minimum air and grass temperature. International journal of biometeorology, 1-12.Smith, L., Liang, Q., James, P., & Lin, W. (2015). Assessing the utility of social media as a data source for flood risk management using a real‐time modelling framework. Journal of Flood Risk Management.Shi, X., Chen, Z., Wang, H., Yeung, D. Y., Wong, W. K., & Woo, W. C. (2015). Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting. arXiv preprint arXiv:1506.04214.Schoen, H., Gayo-Avello, D., Takis Metaxas, P., Mustafaraj, E., Strohmaier, M., & Gloor, P. (2013). The power of prediction with social media. Internet Research, 23(5), 528-543.Xiao, H., Li, Y., Gao, J., Wang, F., Ge, L., Fan, W., & Turaga, D. (2015). Believe it today or tomorrow detecting untrustworthy information from dynamic multi-source data. In Proc. of the SIAM International Conference on Data Mining.You, L., Motta, G., Sacco, D., & Ma, T. (2014, October). Social data analysis framework in cloud and Mobility Analyzer for Smarter Cities. In Service Operations and Logistics, and Informatics (SOLI), 2014 IEEE International Conference on (pp. 96-101). IEEE.
Data Início: 12/11/2015 Hora: 09:00 Data Fim: 12/11/2015 Hora: 14:00
Local: LNCC - Laboratório Nacional de Computação Ciêntifica - Auditorio B
Aluno: Noel Moreno Lemus - - LNCC
Orientador: Fabio Andre Machado Porto - Laboratório Nacional de Computação Científica - LNCC
Participante Banca Examinadora: André da Motta Salles Barreto - GOOGLE - Artur Ziviani - Laboratório Nacional de Computação Científica - LNCC Marco Antonio Casanova - Pontifícia Universidade Católica do Rio de Janeiro - PUC-RIO