关键词： Stock price movement prediction   Machine learning methods   Persistent homology   Technical indicators  

摘要： Predicting direction of stock price movement is notably important to provide a better guidance to assist market participants in making their investment decisions. This study presents a hybrid method combining machine learning methods with persistent homology to improve the prediction performance. Three stock prices namely Kuala Lumpur Composite Index, Kuala Lumpur Stock Exchange Industrial and Kuala Lumpur Stock Exchange Technology sampled from Kuala Lumpur Stock Exchange are selected for experimental evaluation. In particular, persistent homology was applied to obtain a new and useful input vectors of invariant topological features from returns of these stock prices for further classification task using machine learning methods such as logistic regression, artificial neural network, support vector machine and random forest to predict the next day movement direction of Kuala Lumpur Composite Index. For comparative analysis, we compare the proposed method with others, where the machine learning methods are applied independently on stock returns and also on technical indicators respectively. By using the average of prediction performances and pairwise model comparison method, these two evaluation measures revealed that machine learning methods with persistent homology produced better prediction performance. Our results also demonstrated that the combination of support vector machine with persistent homology generates the best outcome. In general, a combination of machine learning methods with persistent homology is an emerging and promising alternative tool for predicting direction of stock price movement. (c) 2020 Elsevier B.V. All rights reserved.

关键词： 二手房   房价预测模型   LightGBM算法   Stacking算法  

摘要： 随着居民生活水平提升，人们对房地产市场各方面信息的关注度也水涨船高，能否精准预测房价变动、科学分析影响因素，不仅能够科学地指导居民购房，也能方便政府进行监督与管控。 随着土地资源的进一步开发利用，二手房在房地产市场占据着愈加重要的地位。作为土地资源紧缺的代表性城市，北京市二手房市场呈现出欣欣向荣的发展趋势，值得我们以其作为典型，进行深入研究。 本文以北京市普通住宅二手房为研究对象，从链家和贝壳二手网站上进行相关数据爬取，选取地铁站口数量、建筑面积、周边幼儿园数量等与区位、建筑和邻里有关的共43个特征变量，试图建立有关北京市二手房价格的相关模型并探索其影响因素。具体来说，选取单一评估模型例如XGBoost、LightGBM、随机森林等，以特征指标为自变量，二手房实际交易价格为因变量构建模型，分别对房价进行评估预测；并进一步构建融合上述模型优点的Stacking组合模型。结果表明，采取组合模型Stacking算法，能够使得样本预测达到最佳。 此外，为进一步分析特征因素对二手房价的影响，本文以成交价格为标准，将二手房分为高、中、低三个等级，结合对比模型中预测效果较好的LightGBM算法，对指标重要性进行综合性度量，分析不同价格二手房的影响指标，并以此给购房者、中介、政府等提供建设性意见。

关键词： Water supply systems   pipe breaks   statistical modelling   machine learning   causal inference  

摘要： Water supply pipes age, deteriorate and break, which puts at risk the continuous provision of safe potable water endangering the public health in cities. Risk management methods are increasingly applied to optimise the capital investment for pipe replacement and rehabilitation, taking into account the probability and hydraulic impact of pipe breaks. As part of this process, however, historic pipe break data and statistical methods should be utilised to gather causal insights for past breaks to inform operational changes and/or capital investment decisions in order to reduce future breaks. This paper presents a comparative study of statistical and machine learning methods to carry out an exploratory causal analysis for historic pipe breaks in an operational water supply network. Regression models for count data and probabilistic models have been developed. The performance of these models was assessed and enhanced with the introduction of interactions and the inclusion of different network representations.

关键词： 机器学习   岩石物理   神经网络   集成学习   电导率  

摘要： 岩芯电导率是地下油气判断及岩石物理分析的重要参数。传统的岩石物理特性分析一般通过对岩芯样本的实验测量分析或对数字岩芯图像数据的有限元数值计算等方法,来获得岩石样本的电导率等宏观物理特性。但这两类分析方法均需耗费较高的人工、时间成本或大量计算资源。随着人工智能等一代数字化分析方法的发展,文中拟开发一种采用机器学习对岩石电特性进行预测的新方法。在研究中,将针对岩芯的三维图像数据使用集成学习(ensemble learning)和人工神经网络(ANN)来预测电导率。其中传统机器学习与多层神经网络的输入特征是几何参数,而三维卷积神经网络输入特征是三维二元分割图像。在研究中比较各种机器学习方法的优劣。实验表明,以三维二元分割图像作输入特征的三维卷积神经网络(3DCNN)比采用Minkowski泛函作输入特征的学习模型在预测岩芯电导率上性能更优。

关键词： 机器学习   露天矿   卡车运行状态   时间预测  

摘要： 针对露天矿卡车从装载点到卸载点的运输循环总时间预测问题,在考虑卡车特征、道路特征和天气特征影响的情况下,通过支持向量机(SVM)、Adaboost和随机森林(RF)3种算法构建了露天矿卡车运行状态时间预测模型。以某大型露天矿山的卡车调度系统采集数据为例,结果表明,针对运行状态时间预测,采用SVM、Adaboost和RF 3种算法预测结果均优于传统平均法,最佳机器学习方法的平均绝对误差降低了81.52%;同时,将运行状态作为行程时间预测的最小单元进行组合预测,比用单一方法预测整个运输过程时间更好,其平均平方误差和平均绝对误差分别相对降低了76.96%和29.90%。

关键词： 降水量   预测   小波分析   集合经验模态分解   机器学习  

摘要： 为准确、可靠地预测安徽省的年降水量,基于安徽省1900~2009年的均一化降水量数据集,使用信号分析技术和机器学习方法建立区域年降水量预测模型。Morlet小波分析和EEMD结果显示,研究区域历史年降水量序列大致存在3、5、20年左右的周期。为提高模型精度,建立5种输入层为3个节点、输出层为1个节点的机器学习模型,即BPNN、WANN、TSNN、SVM、ELM。按4∶1原则,将整理好的样本集中的前85组作为模型训练集,后22组作为测试集。结果表明,5种模型表现较好,率定期的平均相对误差分别为6.1%、12.1%、14.3%、14.3%、13.2%;验证期的平均相对误差为20.6%、13.6%、12.5%、13.0%、14.3%,合格率分别为63.7%、72.7%、77.3%、77.3%、72.7%。总体来看,除BPNN模型外,其余模型均较理想,机器学习方法在非线性水文序列的模拟和预测中具有较好的适用性和可靠性。研究成果可为安徽省未来水资源规划、配置提供指导。

关键词： 价格预测   机器学习   精确度  

摘要： 运用多种机器学习方法对给定电动汽车数据建立了模型,对比发现了逻辑斯蒂回归模型的性能最好,精确度达97.33%,最终选择逻辑斯蒂回归模型用于对电动汽车的价格进行预测。

关键词： Rainfall prediction   Regression   Machine learning   Extremes   Statistical downscaling  

摘要： In this paper, we evaluate the performance of 8 statistical and machine learning methods, driven by atmospheric synoptic patterns, for long-term daily rainfall prediction in a semi-arid climate (Tenerife, Spain). Cross-validation is used to reconstruct 36 years of daily rainfall data at 17 gauges. Prediction is independent for each gauge. The reconstructed series are compared with the observed records in order to select the optimal hyperparameters within each family of models. The predictive performance of the models is evaluated using several metrics and statistics related with rainfall intensity and occurrence at daily, monthly and annual aggregation scales. Multivariate and univariate analysis of variance are used to evaluate the differences among models. The results of our work demonstrate that the performance of most machine learning models is very sensitive to the selected hyperparameters. Neural networks are found to perform best to predict rainfall occurrence and intensity. All methods underestimate the variance of the observed series at daily time scales. Generalized linear models using gamma-distributed errors perform best for predicting rainfall extremes, however, their performance limits its practical applications. Results improve significatively at larger temporal aggregations (monthly or annual) making statistical and machine learning methods more valuable for water resources studies.

关键词： Iterative reconstruction   Low-contrast detectability   Mathematical model observer   Noise texture pattern   Radiation dose  

摘要： Rationale and Objectives To evaluate the performance of iterative reconstruction (IR) and filtered back projection (FBP) images in terms of low-contrast detectability at different radiation doses, IR levels, and slice thickness using the mathematical model observer with a focus on low-contrast *** and Methods The CCT189 MITA CT IQ Low-Contrast Phantom was used and helical scans were performed using a 64-detector CT scanner. Tube voltage was set at 120 kVp and tube current was adjusted from 45 to 600 mA. Images were reconstructed at slice thicknesses of 0.625 and 5.0 mm with FBP and five types of iterative progressive reconstruction with visual modeling (IPV) algorithms. The noise power spectrum (NPS) and normalized NPS were calculated. To evaluate low-contrast detectability, the model observer with the channelized Hotelling observer model was applied using low-contrast modules in the *** The NPS and normalized NPS for IPV images had similar curves as that for FBP images. At a slice thickness of 0.625 mm and equivalent radiation dose level, the mean improvement of low-contrast detectability for IPV images was 1.19-2.15-fold greater than FBP images with corresponding noise reduction levels. At equivalent noise levels of 5.0-8.0 HU, low-contrast detectability of the IPVstd2 to IPVstr2 images as almost the same or better than that of the FBP images. However, the detectability of the IPVstr4 image was lower than that of the FBP image (p = 0.02).Conclusion Low-contrast detectability of the IPV images was improved with a similar normalized NPS as with FBP images. Furthermore, a radiation reduction of >50% was achieved for the IPV images, while maintaining similar low-contrast detectability.

关键词： Three-dimensional displays   Feature extraction   Transforms   Training   Computational modeling   Deep learning   Buildings   Explainable machine learning   point cloud cla-ssification   3D object recognition   computer vision   Saab transform  

摘要： An explainable machine learning method for point cloud classification, called the PointHop method, is proposed in this work. The PointHop method consists of two stages: 1) local-to-global attribute building through iterative one-hop information exchange and 2) classification and ensembles. In the attribute building stage, we address the problem of unordered point cloud data using a space partitioning procedure and developing a robust descriptor that characterizes the relationship between a point and its one-hop neighbor in a PointHop unit. When we put multiple PointHop units in cascade, the attributes of a point will grow by taking its relationship with one-hop neighbor points into account iteratively. Furthermore, to control the rapid dimension growth of the attribute vector associated with a point, we use the Saab transform to reduce the attribute dimension in each PointHop unit. In the classification and ensemble stage, we feed the feature vector obtained from multiple PointHop units to a classifier. We explore ensemble methods to improve the classification performance furthermore. It is shown by experimental results that the PointHop method offers classification performance that is comparable with state-of-the-art methods while demanding much lower training complexity.