针对处理肿瘤基因表达数据特征选择问题,提出了一种特征选择方法 LLE Score.该方法是典型的过滤器类型特征选择方法,在样本类别信息的基础上,LLE Score针对特征向量的局部邻域保存能力进行评价,并且根据评价结果进行特征的选取,以此达...针对处理肿瘤基因表达数据特征选择问题,提出了一种特征选择方法 LLE Score.该方法是典型的过滤器类型特征选择方法,在样本类别信息的基础上,LLE Score针对特征向量的局部邻域保存能力进行评价,并且根据评价结果进行特征的选取,以此达到良好的特征选择效果.在实验部分对肿瘤数据集进行特征选择,并采用支持向量机分类器计算分类准确率.通过分类准确率说明了该方法的有效性.展开更多
We present a methodology for constructing a short-term event risk score in heart failure patients from an ensemble predictor, using bootstrap samples, two different classification rules, logistic regression and linear...We present a methodology for constructing a short-term event risk score in heart failure patients from an ensemble predictor, using bootstrap samples, two different classification rules, logistic regression and linear discriminant analysis for mixed data, continuous or categorical, and random selection of explanatory variables to build individual predictors. We define a measure of the importance of each variable in the score and an event risk measure by an odds-ratio. Moreover, we establish a property of linear discriminant analysis for mixed data. This methodology is applied to EPHESUS trial patients on whom biological, clinical and medical history variables were measured.展开更多
The purpose of this paper is to study the theory of conservative estimating functions in nonlinear regression model with aggregated data. In this model, a quasi-score function with aggregated data is defined. When thi...The purpose of this paper is to study the theory of conservative estimating functions in nonlinear regression model with aggregated data. In this model, a quasi-score function with aggregated data is defined. When this function happens to be conservative, it is projection of the true score function onto a class of estimation functions. By constructing, the potential function for the projected score with aggregated data is obtained, which have some properties of log-likelihood function.展开更多
LightGBM is an open-source, distributed and high-performance GB framework built by Microsoft company. LightGBM has some advantages such as fast learning speed, high parallelism efficiency and high-volume data, and so ...LightGBM is an open-source, distributed and high-performance GB framework built by Microsoft company. LightGBM has some advantages such as fast learning speed, high parallelism efficiency and high-volume data, and so on. Based on the open data set of credit card in Taiwan, five data mining methods, Logistic regression, SVM, neural network, Xgboost and LightGBM, are compared in this paper. The results show that the AUC, F1-Score and the predictive correct ratio of LightGBM are the best, and that of Xgboost is second. It indicates that LightGBM or Xgboost has a good performance in the prediction of categorical response variables and has a good application value in the big data era.展开更多
The traditional threat score based on fixed thresholds for precipitation verification is sensitive to intensity forecast bias. In this study, the neighborhood precipitation threat score is modified by defining the thr...The traditional threat score based on fixed thresholds for precipitation verification is sensitive to intensity forecast bias. In this study, the neighborhood precipitation threat score is modified by defining the thresholds in terms of the percentiles of overall precipitation instead of fixed threshold values. The impact of intensity forecast bias on the calculated threat score is reduced. The method is tested with the forecasts of a tropical storm that re-intensified after making landfall and caused heavy flooding. The forecasts are produced with and without radar data assimilation. The forecast with assimilation of both radial velocity and reflectivity produce precipitation patterns that better match observations but have large positive intensity bias. When using fixed thresholds, the neighborhood threat scores fail to yield high scores for forecasts that have good pattern match with observations, due to large intensity bias. In contrast, the percentile-based neighborhood method yields the highest score for the forecast with the best pattern match and the smallest position error. The percentile-based method also yields scores that are more consistent with object-based verifications, which are less sensitive to intensity bias, demonstrating the potential value of percentile-based verification.展开更多
In the research, the crawler technology was used to evaluate the Taobao silkie, and product evaluation data is the case object. After word segmentation, naive Bayesian, K-means, and TF-IDF related algorithms were used...In the research, the crawler technology was used to evaluate the Taobao silkie, and product evaluation data is the case object. After word segmentation, naive Bayesian, K-means, and TF-IDF related algorithms were used to complete word frequency statistics, sentiment analysis, and cluster analysis. It was found that the quality of rural specialty products was more recognized by e-commerce customers, but its overall emotional score was slightly lower than that of non-origin products. After summing up the drawbacks of the former, targeted countermeasures and suggestions were proposed.展开更多
In this paper, it is discussed that two tests for varying dispersion of binomial data in the framework of nonlinear logistic models with random effects, which are widely used in analyzing longitudinal binomial data. O...In this paper, it is discussed that two tests for varying dispersion of binomial data in the framework of nonlinear logistic models with random effects, which are widely used in analyzing longitudinal binomial data. One is the individual test and power calculation for varying dispersion through testing the randomness of cluster effects, which is extensions of Dean(1992) and Commenges et al (1994). The second test is the composite test for varying dispersion through simultaneously testing the randomness of cluster effects and the equality of random-effect means. The score test statistics are constructed and expressed in simple, easy to use, matrix formulas. The authors illustrate their test methods using the insecticide data (Giltinan, Capizzi & Malani (1988)).展开更多
The present aim is to update, upon arrival of new learning data, the parameters of a score constructed with an ensemble method involving linear discriminant analysis and logistic regression in an online setting, witho...The present aim is to update, upon arrival of new learning data, the parameters of a score constructed with an ensemble method involving linear discriminant analysis and logistic regression in an online setting, without the need to store all of the previously obtained data. Poisson bootstrap and stochastic approximation processes were used with online standardized data to avoid numerical explosions, the convergence of which has been established theoretically. This empirical convergence of online ensemble scores to a reference “batch” score was studied on five different datasets from which data streams were simulated, comparing six different processes to construct the online scores. For each score, 50 replications using a total of 10N observations (N being the size of the dataset) were performed to assess the convergence and the stability of the method, computing the mean and standard deviation of a convergence criterion. A complementary study using 100N observations was also performed. All tested processes on all datasets converged after N iterations, except for one process on one dataset. The best processes were averaged processes using online standardized data and a piecewise constant step-size.展开更多
文摘针对处理肿瘤基因表达数据特征选择问题,提出了一种特征选择方法 LLE Score.该方法是典型的过滤器类型特征选择方法,在样本类别信息的基础上,LLE Score针对特征向量的局部邻域保存能力进行评价,并且根据评价结果进行特征的选取,以此达到良好的特征选择效果.在实验部分对肿瘤数据集进行特征选择,并采用支持向量机分类器计算分类准确率.通过分类准确率说明了该方法的有效性.
文摘We present a methodology for constructing a short-term event risk score in heart failure patients from an ensemble predictor, using bootstrap samples, two different classification rules, logistic regression and linear discriminant analysis for mixed data, continuous or categorical, and random selection of explanatory variables to build individual predictors. We define a measure of the importance of each variable in the score and an event risk measure by an odds-ratio. Moreover, we establish a property of linear discriminant analysis for mixed data. This methodology is applied to EPHESUS trial patients on whom biological, clinical and medical history variables were measured.
文摘The purpose of this paper is to study the theory of conservative estimating functions in nonlinear regression model with aggregated data. In this model, a quasi-score function with aggregated data is defined. When this function happens to be conservative, it is projection of the true score function onto a class of estimation functions. By constructing, the potential function for the projected score with aggregated data is obtained, which have some properties of log-likelihood function.
文摘LightGBM is an open-source, distributed and high-performance GB framework built by Microsoft company. LightGBM has some advantages such as fast learning speed, high parallelism efficiency and high-volume data, and so on. Based on the open data set of credit card in Taiwan, five data mining methods, Logistic regression, SVM, neural network, Xgboost and LightGBM, are compared in this paper. The results show that the AUC, F1-Score and the predictive correct ratio of LightGBM are the best, and that of Xgboost is second. It indicates that LightGBM or Xgboost has a good performance in the prediction of categorical response variables and has a good application value in the big data era.
基金primarily supported by the National 973 Fundamental Research Program of China(Grant No.2013CB430103)the Department of Transportation Federal Aviation Administration(Grant No.NA17RJ1227)through the National Oceanic and Atmospheric Administration+1 种基金supported by the National Science Foundation of China(Grant No.41405100)the Fundamental Research Funds for the Central Universities(Grant No.20620140343)
文摘The traditional threat score based on fixed thresholds for precipitation verification is sensitive to intensity forecast bias. In this study, the neighborhood precipitation threat score is modified by defining the thresholds in terms of the percentiles of overall precipitation instead of fixed threshold values. The impact of intensity forecast bias on the calculated threat score is reduced. The method is tested with the forecasts of a tropical storm that re-intensified after making landfall and caused heavy flooding. The forecasts are produced with and without radar data assimilation. The forecast with assimilation of both radial velocity and reflectivity produce precipitation patterns that better match observations but have large positive intensity bias. When using fixed thresholds, the neighborhood threat scores fail to yield high scores for forecasts that have good pattern match with observations, due to large intensity bias. In contrast, the percentile-based neighborhood method yields the highest score for the forecast with the best pattern match and the smallest position error. The percentile-based method also yields scores that are more consistent with object-based verifications, which are less sensitive to intensity bias, demonstrating the potential value of percentile-based verification.
文摘In the research, the crawler technology was used to evaluate the Taobao silkie, and product evaluation data is the case object. After word segmentation, naive Bayesian, K-means, and TF-IDF related algorithms were used to complete word frequency statistics, sentiment analysis, and cluster analysis. It was found that the quality of rural specialty products was more recognized by e-commerce customers, but its overall emotional score was slightly lower than that of non-origin products. After summing up the drawbacks of the former, targeted countermeasures and suggestions were proposed.
基金The project supported by NNSFC (19631040), NSSFC (04BTJ002) and the grant for post-doctor fellows in SELF.
文摘In this paper, it is discussed that two tests for varying dispersion of binomial data in the framework of nonlinear logistic models with random effects, which are widely used in analyzing longitudinal binomial data. One is the individual test and power calculation for varying dispersion through testing the randomness of cluster effects, which is extensions of Dean(1992) and Commenges et al (1994). The second test is the composite test for varying dispersion through simultaneously testing the randomness of cluster effects and the equality of random-effect means. The score test statistics are constructed and expressed in simple, easy to use, matrix formulas. The authors illustrate their test methods using the insecticide data (Giltinan, Capizzi & Malani (1988)).
文摘The present aim is to update, upon arrival of new learning data, the parameters of a score constructed with an ensemble method involving linear discriminant analysis and logistic regression in an online setting, without the need to store all of the previously obtained data. Poisson bootstrap and stochastic approximation processes were used with online standardized data to avoid numerical explosions, the convergence of which has been established theoretically. This empirical convergence of online ensemble scores to a reference “batch” score was studied on five different datasets from which data streams were simulated, comparing six different processes to construct the online scores. For each score, 50 replications using a total of 10N observations (N being the size of the dataset) were performed to assess the convergence and the stability of the method, computing the mean and standard deviation of a convergence criterion. A complementary study using 100N observations was also performed. All tested processes on all datasets converged after N iterations, except for one process on one dataset. The best processes were averaged processes using online standardized data and a piecewise constant step-size.
