乳腺癌的早期诊断可显著提高其治愈的可能性。近年来,大数据与人工智能技术的蓬勃兴起为乳腺癌在内的多种疾病早期诊断提供技术支持。为提升乳腺癌诊断的准确度,构建基于曲线下面积(area under curve, AUC)改进的Stacking集成模型。首先...乳腺癌的早期诊断可显著提高其治愈的可能性。近年来,大数据与人工智能技术的蓬勃兴起为乳腺癌在内的多种疾病早期诊断提供技术支持。为提升乳腺癌诊断的准确度,构建基于曲线下面积(area under curve, AUC)改进的Stacking集成模型。首先,构建基于v-SVM的AdaBoost集成模型,并将其作为Stacking的元学习器。其次,利用各基学习器的总体AUC值对各基学习器的训练结果进行加权,将加权后的结果作为元学习器的训练集对元学习器进行训练。最后,在WDBC和WBC数据集上进行实证分析。结果表明,基于AUC改进的Stacking集成模型在两个数据集上分别取得较高准确率,可为医生提供更为精细、个性化的诊断依据,进而实现更早介入、更高效治疗的目标。展开更多
采用多层感知器模型、随机森林模型为第一层子模型,极端树模型为第二层元模型,建立基于Stacking集成机器学习的波浪预报算法,并引入邻域平均法抑制在拐点处产生的数值震荡。以长江口外海2016年1-9月的风速和中国近海波高数据为数据源,...采用多层感知器模型、随机森林模型为第一层子模型,极端树模型为第二层元模型,建立基于Stacking集成机器学习的波浪预报算法,并引入邻域平均法抑制在拐点处产生的数值震荡。以长江口外海2016年1-9月的风速和中国近海波高数据为数据源,利用机器学习风速与有效波高之间的关系,将2016年10-11月的风速、波高数据用于预报结果的对比分析,预报前45 d R^2拟合优度达到0.97以上,平均误差最大值为0.08 m,平均相对误差最大值为0.05,预报结果与波浪谱模型结果趋势一致,准确度较高;预报结果后15 d误差增长较快,这与训练集数据中寒潮浪占比较少有关。展开更多
We collected waveforms of a teleseismic event which occurred from January 2007 to October 2011 from 174 broadband seismic stations deployed in Hebei and its adjacent areas. Using the H-k stacking method,the average th...We collected waveforms of a teleseismic event which occurred from January 2007 to October 2011 from 174 broadband seismic stations deployed in Hebei and its adjacent areas. Using the H-k stacking method,the average thickness and Poisson's ratio of the crust are acquired. In order to obtain reliable receiver functions, the broadband seismograms of 488 teleseismic earthquakes occurring in the epicentral distance range from 30° to 90° with magnitudes larger than Mb6. 0 are collected. The results show that crustal thickness have conspicuous lateral heterogeneity and have good correlation to the regional geological tectonic features. Poisson ratio's value is equated with the global models estimates which fluctuate at about 0. 25. Crustal thickness has positive correlation to the topography and the Taihang Mountains form the transition zone of thick and thin crustal thickness. There is an obvious difference in crustal thickness beneath the north and south of the Shanxi earthquake zone and the Poisson ratio of Datong,Ningwu and Anze basins is greater than 0. 3. The crustal thickness beneath the Zhang-Bo( Zhangjiakou-Bohai Sea)earthquake zone decreases from west to north and its Poisson ratio shows conspicuous lateral heterogeneity. The thin crust and low Poisson ratio in the Huabei( North China)basin may correlate with the delamination of the North China craton.展开更多
文摘乳腺癌的早期诊断可显著提高其治愈的可能性。近年来,大数据与人工智能技术的蓬勃兴起为乳腺癌在内的多种疾病早期诊断提供技术支持。为提升乳腺癌诊断的准确度,构建基于曲线下面积(area under curve, AUC)改进的Stacking集成模型。首先,构建基于v-SVM的AdaBoost集成模型,并将其作为Stacking的元学习器。其次,利用各基学习器的总体AUC值对各基学习器的训练结果进行加权,将加权后的结果作为元学习器的训练集对元学习器进行训练。最后,在WDBC和WBC数据集上进行实证分析。结果表明,基于AUC改进的Stacking集成模型在两个数据集上分别取得较高准确率,可为医生提供更为精细、个性化的诊断依据,进而实现更早介入、更高效治疗的目标。
文摘采用多层感知器模型、随机森林模型为第一层子模型,极端树模型为第二层元模型,建立基于Stacking集成机器学习的波浪预报算法,并引入邻域平均法抑制在拐点处产生的数值震荡。以长江口外海2016年1-9月的风速和中国近海波高数据为数据源,利用机器学习风速与有效波高之间的关系,将2016年10-11月的风速、波高数据用于预报结果的对比分析,预报前45 d R^2拟合优度达到0.97以上,平均误差最大值为0.08 m,平均相对误差最大值为0.05,预报结果与波浪谱模型结果趋势一致,准确度较高;预报结果后15 d误差增长较快,这与训练集数据中寒潮浪占比较少有关。
基金funded by the Seismic Situation Tracing Orientation Task,China Earthquake Administration(2014020120)the Science and Technology Support Program of Hebei Province(2014020120)
文摘We collected waveforms of a teleseismic event which occurred from January 2007 to October 2011 from 174 broadband seismic stations deployed in Hebei and its adjacent areas. Using the H-k stacking method,the average thickness and Poisson's ratio of the crust are acquired. In order to obtain reliable receiver functions, the broadband seismograms of 488 teleseismic earthquakes occurring in the epicentral distance range from 30° to 90° with magnitudes larger than Mb6. 0 are collected. The results show that crustal thickness have conspicuous lateral heterogeneity and have good correlation to the regional geological tectonic features. Poisson ratio's value is equated with the global models estimates which fluctuate at about 0. 25. Crustal thickness has positive correlation to the topography and the Taihang Mountains form the transition zone of thick and thin crustal thickness. There is an obvious difference in crustal thickness beneath the north and south of the Shanxi earthquake zone and the Poisson ratio of Datong,Ningwu and Anze basins is greater than 0. 3. The crustal thickness beneath the Zhang-Bo( Zhangjiakou-Bohai Sea)earthquake zone decreases from west to north and its Poisson ratio shows conspicuous lateral heterogeneity. The thin crust and low Poisson ratio in the Huabei( North China)basin may correlate with the delamination of the North China craton.
