Distributed hoeffding trees for pocket data mining

Collaborative mining of distributed data streams in a mobile computing environment is referred to as Pocket Data Mining PDM. Hoeffding trees techniques have been experimentally and analytically validated for data stream classification. In this paper, we have proposed, developed and evaluated the adoption of distributed Hoeffding trees for classifying streaming data in PDM applications. We have identified a realistic scenario in which different users equipped with smart mobile devices run a local Hoeffding tree classifier on a subset of the attributes. Thus, we have investigated the mining of vertically partitioned datasets with possible overlap of attributes, which is the more likely case. Our experimental results have validated the efficiency of our proposed model achieving promising accuracy for real deployment.

Distributed classification for pocket data mining

Distributed and collaborative data stream mining in a mobile computing environment is referred to as Pocket Data Mining PDM. Large amounts of available data streams to which smart phones can subscribe to or sense, coupled with the increasing computational power of handheld devices motivates the development of PDM as a decision making system. This emerging area of study has shown to be feasible in an earlier study using technological enablers of mobile software agents and stream mining techniques [1]. A typical PDM process would start by having mobile agents roam the network to discover relevant data streams and resources. Then other (mobile) agents encapsulating stream mining techniques visit the relevant nodes in the network in order to build evolving data mining models. Finally, a third type of mobile agents roam the network consulting the mining agents for a final collaborative decision, when required by one or more users. In this paper, we propose the use of distributed Hoeffding trees and Naive Bayes classifers in the PDM framework over vertically partitioned data streams. Mobile policing, health monitoring and stock market analysis are among the possible applications of PDM. An extensive experimental study is reported showing the effectiveness of the collaborative data mining with the two classifers.

Homogeneous and heterogeneous distributed classification for pocket data mining

Pocket Data Mining (PDM) describes the full process of analysing data streams in mobile ad hoc distributed environments. Advances in mobile devices like smart phones and tablet computers have made it possible for a wide range of applications to run in such an environment. In this paper, we propose the adoption of data stream classification techniques for PDM. Evident by a thorough experimental study, it has been proved that running heterogeneous/different, or homogeneous/similar data stream classification techniques over vertically partitioned data (data partitioned according to the feature space) results in comparable performance to batch and centralised learning techniques.

Multimodel estimates of intercontinental source-receptor relationships for ozone pollution

Understanding the surface O3 response over a “receptor” region to emission changes over a foreign “source” region is key to evaluating the potential gains from an international approach to abate ozone (O3) pollution. We apply an ensemble of 21 global and hemispheric chemical transport models to estimate the spatial average surface O3 response over east Asia (EA), Europe (EU), North America (NA), and south Asia (SA) to 20% decreases in anthropogenic emissions of the O3 precursors, NOx, NMVOC, and CO (individually and combined), from each of these regions. We find that the ensemble mean surface O3 concentrations in the base case (year 2001) simulation matches available observations throughout the year over EU but overestimates them by >10 ppb during summer and early fall over the eastern United States and Japan. The sum of the O3 responses to NOx, CO, and NMVOC decreases separately is approximately equal to that from a simultaneous reduction of all precursors. We define a continental-scale “import sensitivity” as the ratio of the O3 response to the 20% reductions in foreign versus “domestic” (i.e., over the source region itself) emissions. For example, the combined reduction of emissions from the three foreign regions produces an ensemble spatial mean decrease of 0.6 ppb over EU (0.4 ppb from NA), less than the 0.8 ppb from the reduction of EU emissions, leading to an import sensitivity ratio of 0.7. The ensemble mean surface O3 response to foreign emissions is largest in spring and late fall (0.7–0.9 ppb decrease in all regions from the combined precursor reductions in the three foreign regions), with import sensitivities ranging from 0.5 to 1.1 (responses to domestic emission reductions are 0.8–1.6 ppb). High O3 values are much more sensitive to domestic emissions than to foreign emissions, as indicated by lower import sensitivities of 0.2 to 0.3 during July in EA, EU, and NA when O3 levels are typically highest and by the weaker relative response of annual incidences of daily maximum 8-h average O3 above 60 ppb to emission reductions in a foreign region (<10–20% of that to domestic) as compared to the annual mean response (up to 50% of that to domestic). Applying the ensemble annual mean results to changes in anthropogenic emissions from 1996 to 2002, we estimate a Northern Hemispheric increase in background surface O3 of about 0.1 ppb a−1, at the low end of the 0.1–0.5 ppb a−1 derived from observations. From an additional simulation in which global atmospheric methane was reduced, we infer that 20% reductions in anthropogenic methane emissions from a foreign source region would yield an O3 response in a receptor region that roughly equals that produced by combined 20% reductions of anthropogenic NOx, NMVOC, and CO emissions from the foreign source region.