关键词： Teaching methods   Annual precipitation   Artificial intelligence   Weather extremes   Daily   Precipitation   Neural networks   Climatic extremes   Climate change   Spatial discrimination   Statistical models   Statistical analysis   Machine learning   Climate effects   Disaggregation   Extreme values   Learning algorithms   Statistical modelling   Modelling   Rainfall   Support vector machines   Climate models   Statistical methods   Global climate   Learning   Global climate models   Methodology   Anomalies   Annual   Intercomparison   Mathematical models   Climatic changes  

摘要： Statistical downscaling of Global Climate Models (GCMs) allows researchers to study local climate change effects decades into the future. A wide range of statistical models have been applied to downscaling GCMs but recent advances in machine learning have not been explored compared to traditional approaches. In this paper, we compare five Perfect Prognosis (PP) approaches, Ordinary Least Squares, Elastic-Net, and Support Vector Machine along with two machine learning methods Multi-task Sparse Structure Learning (MSSL) and Autoencoder Neural Networks. In addition, we introduce a hybrid Model Output Statistics and PP approach by modeling the residuals of Bias Correction Spatial Disaggregation (BCSD) with MSSL. Metrics to evaluate each method's ability to capture daily anomalies, large-scale climate shifts, and extremes are analyzed. Generally, we find inconsistent performance between PP methods in their ability to predict daily anomalies and extremes as well as monthly and annual precipitation. However, results suggest that L-1 sparsity constraints aid in reducing error through internal feature selection. The MSSL+BCSD coupling, when compared with BCSD, improved daily, monthly, and annual predictability but decreased performance at the extremes. Hence, these results suggest that the direct application of state-of-the-art machine learning methods to statistical downscaling does not provide direct improvements over simpler, longstanding approaches.

关键词： Machine Learning   CA-125 Antigen   Ovarian Neoplasms   Prognosis   Survival  

摘要： Objectives: The aim of this study was to develop a new prognostic classification for epithelial ovarian cancer (EOC) patients using gradient boosting (GB) and to compare the accuracy of the prognostic model with the conventional statistical method. Methods: Information of EOC patients from Samsung Medical Center (training cohort, n=1,128) was analyzed to optimize the prognostic model using GB. The performance of the final model was externally validated with patient information from Asan Medical Center (validation cohort, n=229). The area under the curve (AUC) by the GB model was compared to that of the conventional Cox proportional hazard regression analysis (CoxPHR) model. Results: In the training cohort, the AUC of the GB model for predicting second year overall survival (OS), with the highest target value, was 0.830 (95% confidence interval [CI]= 0.802-0.853). In the validation cohort, the GB model also showed high AUC of 0.843 (95% CI= 0.833-0.853). In comparison, the conventional CoxPHR method showed lower AUC (0.668 (95% CI= 0.617-0.719) for the training cohort and 0.597 (95% CI= 0.474-0.719) for the validation cohort) compared to GB. New classification according to survival probability scores of the GB model identified four distinct prognostic subgroups that showed more discriminately classified prediction than the International Federation of Gynecology and Obstetrics staging system. Conclusion: Our novel GB-guided classification accurately identified the prognostic subgroups of patients with EOC and showed higher accuracy than the conventional method. This approach would be useful for accurate estimation of individual outcomes of EOC patients.

摘要： Generic drug products are approved by the US Food and Drug Administration (FDA) through Abbreviated New Drug Applications (ANDAs). The ANDA review and approval involves multiple offices across the FDA. Forecasting ANDA submissions can critically inform resource allocation and workload management. In this work, we used machine learning (ML) methodologies to predict the time to first ANDA submissions referencing new chemical entities following their earliest lawful ANDA submission dates. Drug product information, regulatory factors, and pharmacoeconomic factors were used as modeling inputs. The random survival forest ML method, as well as the conventional Cox model, was used for ANDA submission predictions. The ML method outperformed the conventional Cox regression model in predictive performance that was adequately assessed by both internal and external validations. In conclusion, it can potentially serve as an effective forecasting tool for strategic workload and research planning for generic applications.

关键词： Intrusion detection system   Machine learning algorithm   Hybrid system   Feature selection   NSL-KDD  

摘要： In this study, a hybrid and layered Intrusion Detection System (IDS) is proposed that uses a combination of different machine learning and feature selection techniques to provide high performance intrusion detection in different attack types. In the developed system, firstly data preprocessing is performed on the NSL-KDD dataset, then by using different feature selection algorithms, the size of the dataset is reduced. Two new approaches have been proposed for feature selection operation. The layered architecture is created by determining appropriate machine learning algorithms according to attack type. Performance tests such as accuracy, DR, TP Rate, FP Rate, F-Measure, MCC and time of the proposed system are performed on the NSL-KDD dataset. In order to demonstrate the performance of the proposed system, it is compared with the studies in the literature and performance evaluation is done. It has been shown that the proposed system has high accuracy and a low false positive rates in all attack types.

关键词： DNA methylation   Schizophrenia   Machine learning   Classification   Clustering  

摘要： Schizophrenia is a common mental disorder with high heritability. It is genetically complex and to date more than a hundred risk loci have been identified. Association of environmental factors and schizophrenia has also been reported, while epigenetic analyses have yielded ambiguous and sometimes conflicting results. Here, we analyzed fresh frozen post-mortem brain tissue from a cohort of 73 subjects diagnosed with schizophrenia and 52 control samples, using the Illumina Infinium HumanMethylation450 Bead Chip, to investigate genome-wide DNA methylation patterns in the two groups. Analysis of differential methylation was performed with the Bioconductor Minfi package and modern machine-learning and visualization techniques, which were shown previously to be successful in detecting and highlighting differentially methylated patterns in case-control studies. In this dataset, however, these methods did not uncover any significant signals discerning the patient group and healthy controls, suggesting that if there are methylation changes associated with schizophrenia, they are heterogeneous and complex with small effect.

关键词： Alzheimer's disease   ADNI   Cortical thickness   Gyrification index   Fractals   ADAS cognitive test   Machine learning   Classification   Support vector machine  

摘要： Early detection of Alzheimer's disease (AD) using structural magnetic resonance images is essential for early treatment that can slow the progression of the disease. Therefore, there is a need for accurate computer-aided-diagnosis (CAD) systems for detecting AD. The purpose of this work is to evaluate the degree to which specific features - including fractals obtained from MRI-based surfaces of the cerebral cortex, cortical thickness, gyrification index and the Alzheimer's disease assessment scale (ADAS) cognitive test scores - are informative for classifying AD patients and healthy control subjects using several machine learning classifiers. Our results show that a Support Vector Machine (SVM) trained with cortical thickness, gyrification index and ADAS cognitive test scores distinguishes between AD and healthy control subjects better than other machine learning methods and other feature combinations. This specific CAD system achieved ideal accuracy and outperformed recently proposed systems. (C) 2018 Published by Elsevier Ltd.

关键词： Feature extraction   Machine learning   Classification   Mammography   Breast cancer   Gabor wavelet  

摘要： Among the major causes of female mortality, breast cancer used to pose big challenges to the medical world. Currently, the most popular method of monitoring and diagnosesin addition to mammographyis carrying out repeated biopsies to locate the tumor further, that may result in loss of breast tissues. This paper presents an effective method of classifying and detecting the masses in mammograms. In the proposed method, we first attain the feature vector pertaining to each mammography image based on Gabor wavelet transform. Then, we performed tenfold cross validation through several experiments, analyzing the data complexity on each fold. We also used some machine learning methods as decision-making stage and achieved mean accuracies above 0.939, mean sensitivities as high as 0.951, and the mean specificities greater than 0.92. Evaluations and comparisons witness the effectiveness of the proposed method for better diagnosis of breast cancer against the known classification techniques developed in mammography. Simplicity, robustness and high accuracy are advantages of the proposed method.

关键词： Actinobacterial strains   Colony features   Image processing   Deep neural network   Transfer learning   ResNet  

摘要： Recognition of actinobacterial species on solid culture plates is an error-prone and time-consuming process which retard all the down-stream process after the isolation step. In this paper, we propose an optimized solution for the mentioned problem by using machine learning and image processing algorithms to diminish the cost and time and increase the accuracy of the detection. Three methods are compared in this paper for actinobacterial strains recognition. In the first method, two-level wavelet transform is applied on images of actinobacterial strains and statistical texture features are computed from wavelet subbands. Furthermore, some statistical color features are calculated from color information. In consequence, Principle Component Analysis (PCA) is employed for dimension reduction and finally, a Multi-Layer Perceptron (MLP) neural network was used for classification. In the second method, a Convolution Neural Network (CNN) is used to extract the features automatically. In the third method, transfer learning is employed for feature extraction and classfication. The first method is evaluated on two databases, *** and ***, and the accuracy obtained between 80.8% to 80.1%, respectively. Employing CNN as the feature extractor improved the accuracy about 4%. Transfer learnig results 85.96% accuracy on the *** and 91.90% on the subclasses of ***. The experiments have shown that using transfer learning of DCNN of type ResNet has better performance compared to the pervious methods. The proposed methods are universal and can be used for recognition of other circle-like colony-shape microorganisms. In particular, giving limited and unbalanced training data, which is a common failure in biological data sets, the proposed methods harbor remarkable accuracy. The data augmentation methods showed to be efficient and practical for the current purpose along with being easy to be implemented and integrated.

关键词： Pixels  

摘要： The problem of detecting neurons in optical microscopy is considered by the example of Nissl-stained mouse brain slices. The proposed algorithm consists of the following steps: preprocessing, textural feature extraction, pixel classification, and pixel clustering. When solving this problem, we investigate various preprocessing methods, machine learning algorithms, and textural features. At the classification step, the k nearest neighbor (kNN) algorithm or the naive Bayes classifier (NBC) is used to determine whether each pixel of the image belongs to the neuron's cell body. In this paper, we investigate textural features of two types: features based on the normalized histogram and features based on the gray level co-occurrence matrix (GLCM). To find the centers of the neurons, all pixels that belong to the neurons are clustered using the mean shift algorithm. It is shown that the best detection quality (precision = 0.82, recall = 0.92, and F1 = 0.86) is achieved with GLCM-based features and neighborhood radius R = 6. It is also shown that the selection of different preprocessing algorithms significantly affects the detection result. In terms of detection quality, the kNN algorithm outperforms the NBC. On the dataset used, the selection of the parameter k > 15 does not significantly improve the quality of detection. The proposed method yields the result similar to that achieved in [1]: Recall(A) = 865%. In sampling tests on some microscopy images from the Broad Bioimage Benchmark Collection (BBBC), the proposed approach shows the best or equivalent quality in detecting the number of cells on the image. For detection, the algorithm uses only local textural features, which removes restrictions on the parallelization of computations.

关键词： machine learning   rapid damage assessment   seismic instrumentation and monitoring   seismic response demands   tall building clusters  

摘要： An approach to reconstructing full-profile seismic response demands across multiple tall buildings, using kernel-based machine learning methods, is introduced. Nonlinear response history analyses are used to generate a dataset of peak floor accelerations and peak story drift ratios for a portfolio of tall buildings, using spatially explicit ground motions from the Northridge earthquake. Structural dissimilarities are incorporated by including a range of building heights and differences in the type and combination of lateral force resisting systems. Using measurements from limited locations within a subset of buildings, the full-profile response demands for all buildings in a portfolio are reconstructed. A rigorous evaluation procedure is used to demonstrate the ability of the kernel-based methods to accurately capture the highly nonlinear response demand patterns within and across buildings. For a scenario where the first floor, mid-height, and roof level responses are known for 40% of the buildings, the kernel-based machine learning methods are able to estimate the full-profile demands of the entire portfolio with a median error that is approximately 30% of the measured demands.