Data-driven models for monthly streamflow time series prediction

Data-driven techniques such as Auto-Regressive Moving Average (ARMA), K-Nearest-Neighbors (KNN), and Artificial Neural Networks (ANN), are widely applied to hydrologic time series prediction. This paper investigates different data-driven models to determine the optimal approach of predicting monthly streamflow time series. Four sets of data from different locations of People’s Republic of China (Xiangjiaba, Cuntan, Manwan, and Danjiangkou) are applied for the investigation process. Correlation integral and False Nearest Neighbors (FNN) are first employed for Phase Space Reconstruction (PSR). Four models, ARMA, ANN, KNN, and Phase Space Reconstruction-based Artificial Neural Networks (ANN-PSR) are then compared by one-month-ahead forecast using Cuntan and Danjiangkou data. The KNN model performs the best among the four models, but only exhibits weak superiority to ARMA. Further analysis demonstrates that a low correlation between model inputs and outputs could be the main reason to restrict the power of ANN. A Moving Average Artificial Neural Networks (MA-ANN), using the moving average of streamflow series as inputs, is also proposed in this study. The results show that the MA-ANN has a significant improvement on the forecast accuracy compared with the original four models. This is mainly due to the improvement of correlation between inputs and outputs depending on the moving average operation. The optimal memory lengths of the moving average were three and six for Cuntan and Danjiangkou, respectively, when the optimal model inputs are recognized as the previous twelve months.     

Trend Autoregressive Model Exact Run Length Evaluation on a Two-Sided Extended EWMA Chart

The Extended Exponentially Weighted Moving Average (extended EWMA) control chart is one of the control charts and can be used to quickly detect a small shift. The performance of control charts can be evaluated with the average run length (ARL). Due to the deriving explicit formulas for the ARL on a two-sided extended EWMA control chart for trend autoregressive or trend AR(p) model has not been reported previously. The aim of this study is to derive the explicit formulas for the ARL on a two-sided extended EWMA control chart for the trend AR(p) model as well as the trend AR(1) and trend AR(2) models with exponential white noise. The analytical solution accuracy was obtained with the extended EWMA control chart and was compared to the numerical integral equation (NIE) method. The results show that the ARL obtained by the explicit formula and the NIE method is hardly different, but the explicit formula can help decrease the computational (CPU) time. Furthermore, this is also expanded to comparative performance with the Exponentially Weighted Moving Average (EWMA) control chart. The performance of the extended EWMA control chart is better than the EWMA control chart for all situations, both the trend AR(1) and trend AR(2) models. Finally, the analytical solution of ARL is applied to real-world data in the health field, such as COVID-19 data in the United Kingdom and Sweden, to demonstrate the efficacy of the proposed method.     

CBOE Volatility Index Forecasting under COVID-19: An Integrated BiLSTM-ARIMA-GARCH Model

After the outbreak of COVID-19, the global economy entered a deep freeze. This observation is supported by the Volatility Index (VIX), which reflects the market risk expected by investors. In the current study, we predicted the VIX using variables obtained from the sentiment analysis of data on Twitter posts related to the keyword “COVID-19,” using a model integrating the bidirectional long-term memory (BiLSTM), autoregressive integrated moving average (ARIMA) algorithm, and generalized autoregressive conditional heteroskedasticity (GARCH) model. The Linguistic Inquiry and Word Count (LIWC) program and Valence Aware Dictionary for Sentiment Reasoning (VADER) model were utilized as sentiment analysis methods. The results revealed that during COVID-19, the proposed integrated model, which trained both the Twitter sentiment values and historical VIX values, presented better results in forecasting the VIX in time-series regression and direction prediction than those of the other existing models.     

基于预训练-微调策略的COVID-19预测模型

COVID-19的世界性大流行对整个社会产生了严重的影响,通过数学建模对确诊病例数进行预测将有助于为公共卫生决策提供依据。在复杂多变的外部环境下,基于深度学习的传染病预测模型成为研究热点。然而,现有模型对数据量要求较高,在进行监督学习时不能很好地适应低数据量的场景,导致预测精度降低。构建结合预训练-微调策略的COVID-19预测模型P-GRU。通过在源地区数据集上采用预训练策略,使模型提前获得更多的疫情数据,从而学习到COVID-19的隐式演变规律,为模型预测提供更充分的先验知识,同时使用包含最近历史信息的固定长度序列预测后续时间点的确诊病例数,并在预测过程中考虑本地人为限制政策因素对疫情趋势的影响,实现针对目标地区数据集的精准预测。实验结果表明,预训练策略能够有效提高预测性能,相比于卷积神经网络、循环神经网络、长短期记忆网络和门控循环单元模型,P-GRU模型在平均绝对百分比误差和均方根误差评价指标上表现优异,更适合用于预测COVID-19传播趋势。     

Experimental study and Random Forest prediction model of microbiome cell surface hydrophobicity

The cell surface hydrophobicity (CSH) is an assessable physicochemical property used to evaluate the microbial adhesion to the surface of biomaterials, which is an essential step in the microbial biofilm formation and pathogenesis. For the present in vitro fermentation experiment, the CSH of ruminal mixed microbes was considered, along with other data records of pH, ammonia-nitrogen concentration, and neutral detergent fibre digestibility, conditions of surface tension and specific surface area in two different time scales. A dataset of 170,707 perturbations of input variables, grouped into two blocks of data, was constructed. Next, Expected Measurement Moving Average – Machine Learning (EMMA-ML) models were developed in order to predict CSH after perturbations of all input variables. EMMA-ML is a Perturbation Theory method that combines the ideas of Expected Measurement, Box-Jenkins Operators/Moving Average, and Time Series Analysis. Seven regression methods have been tested: Multiple Linear regression, Generalized Linear Model with Stepwise Feature Selection, Partial Least Squares regression, Lasso regression, Elastic Net regression, Neural Networks regression, and Random Forests (RF). The best regression performance has been obtained with RF (EMMA-RF model) with an R-squared of 0.992. The model analysis has shown that CSH values were highly dependent on the in vitro fermentation parameters of detergent fibre digestibility, ammonia – nitrogen concentration, and the expected values of cell surface hydrophobicity in the first time scale.     

基于ARIMA与SVM的国际铀资源价格预测

由于国际铀资源价格时间序列数据的非线性性与非平稳性,使用单一的预测模型很难捕捉到其综合趋势。为了进一步提高模型的预测精度,建立了基于差分自回归移动平均（ARIMA）和支持向量机SVM的组合预测模型,并用PSO算法对SVM模型中的参数进行优化。将该方法应用于实际铀资源价格预测,并与单一的ARIMA模型和SVM模型进行比较。仿真实验结果表明,该组合预测模型实现了对铀资源价格数据更为准确的预测。     

Dynamic Ensemble Multivariate Time Series Forecasting Model for PM2.5

In forecasting real time environmental factors, large data is needed to analyse the pattern behind the data values. Air pollution is a major threat towards developing countries and it is proliferating every year. Many methods in time series prediction and deep learning models to estimate the severity of air pollution. Each independent variable contributing towards pollution is necessary to analyse the trend behind the air pollution in that particular locality. This approach selects multivariate time series and coalesce a real time updatable autoregressive model to forecast Particulate matter (PM) PM2.5. To perform experimental analysis the data from the Central Pollution Control Board (CPCB) is used. Prediction is carried out for Chennai with seven locations and estimated PM’s using the weighted ensemble method. Proposed method for air pollution prediction unveiled effective and moored performance in long term prediction. Dynamic budge with high weighted k-models are used simultaneously and devising an ensemble helps to achieve stable forecasting. Computational time of ensemble decreases with parallel processing in each sub model. Weighted ensemble model shows high performance in long term prediction when compared to the traditional time series models like Vector Auto-Regression (VAR), Autoregressive Integrated with Moving Average (ARIMA), Autoregressive Moving Average with Extended terms (ARMEX). Evaluation metrics like Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and the time to achieve the time series are compared.     

GPRS小区流量预测中时序模型的比较研究

针对通用无线分组业务(GPRS)小区流量预测问题,对几种典型时序预测模型的性能进行了综合分析。在总结时序预测模型使用步骤的基础上,分析了自回归（AR）、自回归移动平均（ARIMA）和乘积季节自回归求和移动平均（ARIMA）模型的性能。首先,对GPRS小区流量的变化情况进行分析;再根据流量的自相关系数和偏相关系数,从不同的角度进行分析,分别得到了流量变化的AR模型和ARMA模型;进而利用小区流量以天为周期变化的特点,得到了流量变化的乘积季节ARIMA模型。最后根据GPRS小区历史流量数据,应用这三种模型预测将来某一时间的流量,并对模型性能进行比较研究。     

基于构造性神经网络的时间序列混合预测模型 *

针对传统时间序列预测模型不适应非线性预测而适应非线性预测的BP算法存在收敛速度慢,且容易陷入局部极小等问题,提出一种基于构造性神经网络的时间序列混合预测模型。采用构造性神经网络模型（覆盖算法）得出的类别值对统计时间序列模型的预测值进行修正,建立一种同时考虑时间序列自身周期变化和外生变量因子对时间序列未来变化趋势影响的混合预测模型,涵盖了实际问题的线性和非线性两方面,提高了预测精度。将该模型应用到粮食产量的预测中,取得了较好的预测效果。     

Modeling MODIS LAI time series using three statistical methods

Leaf Area Index (LAI) is one of the most important variables characterizing land surface vegetation and dynamics. Many satellite data, such as the Moderate Resolution Imaging Spectroradiometer (MODIS), have been used to generate LAI products. It is important to characterize their spatial and temporal variations by developing mathematical models from these products. In this study, we aim to model MODIS LAI time series and further predict its future values by decomposing the LAI time series of each pixel into several components: trend, intra-annual variations, seasonal cycle, and stochastic stationary or irregular parts. Three such models that can characterize the non-stationary time series data and predict the future values are explored, including Dynamic Harmonics Regression (DHR), STL (Seasonal-Trend Decomposition Procedure based on Loess), and Seasonal ARIMA (AutoRegressive Intergrated Moving Average) (SARIMA). The preliminary results using six years (2001-2006) of the MODIS LAI product indicate that all these methods are effective to model LAI time series and predict 2007 LAI values reasonably well. The SARIMA model gives the best prediction, DHR produces the smoothest curve, and STL is more sensitive to noise in the data. These methods work best for land cover types with pronounced seasonal variations.     

基于异构并行神经网络的语音情感识别

提取能表征语音情感的特征并构建具有较强鲁棒性和泛化性的声学模型是语音情感识别系统的核心。面向语音情感识别构建基于注意力机制的异构并行卷积神经网络模型AHPCL,采用长短时记忆网络提取语音情感的时间序列特征,使用卷积操作提取语音空间谱特征,通过将时间信息和空间信息相结合共同表征语音情感,提高预测结果的准确率。利用注意力机制,根据不同时间序列特征对语音情感的贡献程度分配权重,实现从大量特征信息中选择出更能表征语音情感的时间序列。在CASIA、EMODB、SAVEE等3个语音情感数据库上提取音高、过零率、梅尔频率倒谱系数等低级描述符特征,并计算这些低级描述符特征的高级统计函数共得到219维的特征作为输入进行实验验证。结果表明,AHPCL模型在3个语音情感数据库上分别取得了86.02%、84.03%、64.06%的未加权平均召回率,相比LeNet、DNN-ELM和TSFFCNN基线模型具有更强的鲁棒性和泛化性。     

基于构造性神经网络的时间序列混合预测模型 *

针对传统时间序列预测模型不适应非线性预测而适应非线性预测的 BP算法存在收敛速度慢 ,且容易陷入局部极小等问题 ,提出一种基于构造性神经网络的时间序列混合预测模型。采用构造性神经网络模型 (覆盖算法 )得出的类别值对统计时间序列模型的预测值进行修正 ,建立一种同时考虑时间序列自身周期变化和外生变量因子对时间序列未来变化趋势影响的混合预测模型 ,涵盖了实际问题的线性和非线性两方面 ,提高了预测精度。将该模型应用到粮食产量的预测中 ,取得了较好的预测效果。     

基于Lasso-MIDAS模型的混频时间序列预测研究

针对传统的时间序列预测方法在处理复杂丰富的大数据时常面临变量间抽样频率不同、数据相关性复杂等问题,基于Lasso算法和混频数据抽样模型(MIDAS)提出了不改变数据结构的混频时序预测模型Lasso-MIDAS。该模型通过融合MIDAS处理混频信息的机制和Lasso算法的压缩特性来实现估计预测,实时修正对预测最有效的混频变量集;根据常见的正则化方法岭回归设计了Ridge-MIDAS模型用做对比。实验结果表明,Lasso-MIDAS在预测性能上优于标准MIDAS模型及对比模型,验证了该方法在混频时间序列预测方面的有效性。     

融合微博情感分析和深度学习的宏观经济预测方法

现代市场经济快速发展的同时也伴随着较高的风险,通过对地区投资情况提前预测,能够提前发现投资风险,为国家、企业的投资决策提供参考。针对宏观经济预测中统计数据滞后和内部关系复杂的问题,提出融合情感分析和深度学习的预测方法（SA-LSTM）。首先考虑微博的强时效性,确定了微博爬取和情感分析的方法,得到微博情感分析的分值,进而结合政府统计的结构化经济指标和长短期记忆神经网络,实现地区投资总额预测。经过实际数据计算验证,在四个数据集上,与不加入微博情感分析的LSTM网络相比,SA-LSTM能够降低预测相对误差4.95,0.92,1.21,0.66个百分点;与差分自回归移动平均模型（ARIMA）、线性回归（LR）、反向传播（BP）神经网络、长短期记忆（LSTM）网络四个方法中的最优方法相比能够降低相对误差0.06,0.92,0.94,0.66个百分点。另外,SA-LSTM在多个时间片上,预测相对误差的方差最小,表明所提方法具有很好的鲁棒性,对数据抖动有良好的适应性。     

基于组合模型的自相似业务流量预测

针对经验模式分解存在的模态混叠问题,提出了一种基于组合模型的自相似业务流量预测方法。首先通过对网络流量进行集合经验模式分解,有效地去除自相似网络流量中存在的长相关性。接着根据分解得到的各本征模态函数分量的不同特性,分别采用人工神经网络与自回归滑动平均模型对其进行预测,最终再将预测结果进行组合。仿真结果表明,提出的方法对于实际网络流量数据具有较高的预测精度。     

Predictive accuracy comparison between neural networks and statistical regression for development effort of software projects

To get a better prediction of costs, schedule, and the risks of a software project, it is necessary to have a more accurate prediction of its development effort. Among the main prediction techniques are those based on mathematical models, such as statistical regressions or machine learning (ML). The ML models applied to predicting the development effort have mainly based their conclusions on the following weaknesses: (1) using an accuracy criterion which leads to asymmetry, (2) applying a validation method that causes a conclusion instability by randomly selecting the samples for training and testing the models, (3) omitting the explanation of how the parameters for the neural networks were determined, (4) generating conclusions from models that were not trained and tested from mutually exclusive data sets, (5) omitting an analysis of the dependence, variance and normality of data for selecting the suitable statistical test for comparing the accuracies among models, and (6) reporting results without showing a statistically significant difference. In this study, these six issues are addressed when comparing the prediction accuracy of a radial Basis Function Neural Network (RBFNN) with that of a regression statistical (the model most frequently compared with ML models), to feedforward multilayer perceptron (MLP, the most commonly used in the effort prediction of software projects), and to general regression neural network (GRNN, a RBFNN variant). The hypothesis tested is the following: the accuracy of effort prediction for RBFNN is statistically better than the accuracy obtained from a simple linear regression (SLR), MLP and GRNN when adjusted function points data, obtained from software projects, is used as the independent variable. Samples obtained from the International Software Benchmarking Standards Group (ISBSG) Release 11 related to new and enhanced projects were used. The models were trained and tested from a leave-one-out cross-validation method. The criteria for evaluating the models were based on Absolute Residuals and by a Friedman statistical test. The results showed that there was a statistically significant difference in the accuracy among the four models for new projects, but not for enhanced projects. Regarding new projects, the accuracy for RBFNN was better than for a SLR at the 99% confidence level, whereas the MLP and GRNN were better than for a SLR at the 90% confidence level.     

新冠肺炎疫情趋势预测模型

2019年新型冠状病毒肺炎(corona virus disease 2019,COVID-19)的爆发对人们的健康和生活造成了极大的危害和影响。预测疫情的发展趋势可帮助人们提前制定应对措施。SEIR模型是经典的传染病模型之一,由于该模型中病毒传染率为常数,难以对新冠肺炎传播情况进行准确建模并完成疫情趋势预测。针对此问题,本文提出基于长短期记忆网络(long short-term memory,LSTM)的病毒传染率预测方法,并将其与SEIR模型结合,建立新冠肺炎疫情趋势预测模型(LSTM-SEIR network, LS-Net)。为了验证本文提出的方法,收集了国内多个省市官方公布的疫情数据进行实验。实验结果表明,本文提出的LS-Net可对疫情发展趋势进行有效预测,并优于传统SEIR模型。     

一种新的统计预测模型——多项式系数自回归模型

传统的自回归滑动平均模型（ARMA）和新近出现的函数系数自回归模型（FAR）不能满足非线性时间序列预测分析的准确度与运算速度要求,为了改进预测性能,研究提出了一种新的统计预测模型——多项式系数自回归模型（PCAR）。给出了PCAR模型的表示形式,详细探讨了PCAR模型的参数估计和阶次选择方法,在此基础上又提出了基于BIC准则的建模算法。同ARMA模型相比,PCAR模型扩大了适用对象范围,有效降低了模型选择误差;同FAR模型相比,它具有参数模型的特点,避免了系数函数局部线性回归估计所存在的不足;分析了PCAR模型与ARMA、FAR模型的等价条件。通过实验分析得出了PCAR模型较ARMA、FAR模型的单步预测准确度分别提高了99.65%和18.7%的结论,而且PCAR建模运算所需时间仅为FAR模型的0.2%。