针对2007年6月—9月全球集合预报系统(THORPEX Interactive Grand Global Ensemble,TIGGE)多模式中的欧洲中期天气预报中心(European Centre for Medium-Range Weather Forecasts,ECMWF)、美国国家环境预报中心(National Centers for En...针对2007年6月—9月全球集合预报系统(THORPEX Interactive Grand Global Ensemble,TIGGE)多模式中的欧洲中期天气预报中心(European Centre for Medium-Range Weather Forecasts,ECMWF)、美国国家环境预报中心(National Centers for Environmental Prediction,NCEP)、英国气象局(United Kingdom Meteorological Office,UKMO)及日本气象厅(Japan Meteorological Agency,JMA)这4个集合预报模式产品资料,对东亚地区地面2 m气温进行了贝叶斯模式平均(Bayesian model averaging,BMA)方法的概率预报研究,并与简单集合平均(ensemble mean,EMN)方法的概率预报进行对比,最终采用(continuous ranked probability score,CRPS)评分与均方根误差(root of mean square error,RMSE)评估BMA方法在概率预报和确定性预报方面的预报效果。研究表明,BMA方法可有效将单一的确定性预报向连续概率预报转化,并全面准确地描述大气的多种可能运动状态,同时提供概率预报和确定性预报。BMA方法在量化不确定性信息方面优于EMN方法,在一定程度上减小了预报的不确定性。就确定性预报效果而言,BMA方法的预报效果优于所有的单模式预报以及EMN方法,但会受到训练期长度和选取模式性能优劣的影响,其预报效果也会发生改变。展开更多
In the privacy preservation of association rules, sensitivity analysis should be reported after the quantification of items in terms of their occurrence. The traditional methodologies, used for preserving confidential...In the privacy preservation of association rules, sensitivity analysis should be reported after the quantification of items in terms of their occurrence. The traditional methodologies, used for preserving confidentiality of association rules, are based on the assumptions while safeguarding susceptible information rather than recognition of insightful items. Therefore, it is time to go one step ahead in order to remove such assumptions in the protection of responsive information especially in XML association rule mining. Thus, we focus on this central and highly researched area in terms of generating XML association rule mining without arguing on the disclosure risks involvement in such mining process. Hence, we described the identification of susceptible items in order to hide the confidential information through a supervised learning technique. These susceptible items show the high dependency on other items that are measured in terms of statistical significance with Bayesian Network. Thus, we proposed two methodologies based on items probabilistic occurrence and mode of items. Additionally, all this information is modeled and named PPDM (Privacy Preservation in Data Mining) model for XARs. Furthermore, the PPDM model is helpful for sharing markets information among competitors with a lower chance of generating monopoly. Finally, PPDM model introduces great accuracy in computing sensitivity of items and opens new dimensions to the academia for the standardization of such NP-hard problems.展开更多
文摘针对2007年6月—9月全球集合预报系统(THORPEX Interactive Grand Global Ensemble,TIGGE)多模式中的欧洲中期天气预报中心(European Centre for Medium-Range Weather Forecasts,ECMWF)、美国国家环境预报中心(National Centers for Environmental Prediction,NCEP)、英国气象局(United Kingdom Meteorological Office,UKMO)及日本气象厅(Japan Meteorological Agency,JMA)这4个集合预报模式产品资料,对东亚地区地面2 m气温进行了贝叶斯模式平均(Bayesian model averaging,BMA)方法的概率预报研究,并与简单集合平均(ensemble mean,EMN)方法的概率预报进行对比,最终采用(continuous ranked probability score,CRPS)评分与均方根误差(root of mean square error,RMSE)评估BMA方法在概率预报和确定性预报方面的预报效果。研究表明,BMA方法可有效将单一的确定性预报向连续概率预报转化,并全面准确地描述大气的多种可能运动状态,同时提供概率预报和确定性预报。BMA方法在量化不确定性信息方面优于EMN方法,在一定程度上减小了预报的不确定性。就确定性预报效果而言,BMA方法的预报效果优于所有的单模式预报以及EMN方法,但会受到训练期长度和选取模式性能优劣的影响,其预报效果也会发生改变。
文摘In the privacy preservation of association rules, sensitivity analysis should be reported after the quantification of items in terms of their occurrence. The traditional methodologies, used for preserving confidentiality of association rules, are based on the assumptions while safeguarding susceptible information rather than recognition of insightful items. Therefore, it is time to go one step ahead in order to remove such assumptions in the protection of responsive information especially in XML association rule mining. Thus, we focus on this central and highly researched area in terms of generating XML association rule mining without arguing on the disclosure risks involvement in such mining process. Hence, we described the identification of susceptible items in order to hide the confidential information through a supervised learning technique. These susceptible items show the high dependency on other items that are measured in terms of statistical significance with Bayesian Network. Thus, we proposed two methodologies based on items probabilistic occurrence and mode of items. Additionally, all this information is modeled and named PPDM (Privacy Preservation in Data Mining) model for XARs. Furthermore, the PPDM model is helpful for sharing markets information among competitors with a lower chance of generating monopoly. Finally, PPDM model introduces great accuracy in computing sensitivity of items and opens new dimensions to the academia for the standardization of such NP-hard problems.
