Comparative study on the performance of ConvLSTM and ConvGRU in classification problems-taking early warning of short-duration heavy rainfall as an example

卷积长短期记忆单元ConvLSTM和卷积门控循环单元ConvGRU是两种广泛应用的深度学习单元,通过将循环机制与卷积运算相结合,常常用于时空序列的预测.为了明确上述两种模型的收敛速度和分类能力,需要使用相同的模型架构对相同的分类问题进行预测.本研究将北京短时强降水区级预警问题看作深度学习中的二分类问题,使用京津冀雷达网的组合反射率数据和北京区域内的自动气象站降雨数据进行深度学习模型的训练和评估.结果表明,ConvGRU的收敛速度比ConvLSTM快约25%.ConvLSTM和ConvGRU的预警性能随地区,时间,降雨强度的变化趋势相似,但大部分ConvLSTM的得分较高,少数情况下ConvGRU的得分较高.

Effectiveness of Short Message Service Support on Adherence to Chemotherapy Treatment among Patients Attending Cancer Treatment Facilities in Kenya

Introduction: Cancer is a chronic debilitating disease that unnerves patients, communities, and nations. At some point in cancer patient’s disease experience, chemotherapy is used, and the patient is expected to adhere to treatment to improve survival and quality of life. Methods: This multisite Cluster Randomized Trial (CRT) evaluated the effectiveness of mobile phone Short Message Service (SMS) support on the adherence to treatment schedules among adult cancer patients in Kenya. Data was collected using questionnaires. Ethical approvals were obtained from relevant Ethical Review Boards (ERBs). Results: The mean adherence was 83%. There was a significant difference between treatment arms in relation to the adherence. The intervention arm had a higher mean adherence difference, M = 3.913, 95% CI 2.632-5.193, t (402) = 6.006, p ≤ 0.001), with Cohen’s d = 0.60. Although not significant, (χ2dd = 0.151, df = 1, p = 2.064), more women were perfect adheres than males. Perfect adherers were satisfied with SMS support (χ2dd = 7.620, df = 1, p = 0.06), were in the intervention arm (χ2dd = 22.942, df = 1, p ≤ 0.001), and had trust in the care provider (χ2dd = 10.591 p ≤ 0.001). SMS support was not significant in the multivariate analysis but had an estimated effect size of 0.958 (z = 1.424, p = 0.154, CI = 0.242-3.781), indicating that mean adherence was slightly better in the presence of the intervention. Conclusions: SMS-support intervention has demonstrated superiority in influencing adherence. Further, health system-related factors have a significant influence on the adherence to chemotherapy treatment. Interventions to re-design health systems that are responsive to unmet care needs of cancer patients must be explored. .

Off-Pump Coronary Artery Bypass Grafting in Patients with Left Ventricular Dysfunction: Short-Term Results from a Single Center in Bangladesh

Background: Off-pump coronary artery bypass grafting (OPCAB) is considered a safer alternative to on-pump surgery, especially in patients with left ventricular dysfunction (LVD). Objectives: This study assessed short-term outcomes and functional improvements in LVD patients post-OPCAB. Methods: The study included 200 coronary artery disease patients who underwent isolated off-pump coronary artery bypass grafting (OPCAB) at the National Heart Foundation Hospital and Research Institute between January 2019 and June 2020. Patients were categorized into Group 1, with a left ventricular ejection fraction (LVEF) of 30% - 39%, and Group 2, with an LVEF of 40% or higher. Echocardiographic assessments of left ventricular dimensions and ejection fraction were performed preoperatively, at discharge, and one month postoperatively. Results: In Group 1, preoperative left ventricular internal dimensions during diastole (LVIDd) and systole (LVIDs) were 53.48 ± 4.40 mm and 44.23 ± 3.93 mm, respectively, with a left ventricular ejection fraction (LVEF) of 35.28% ± 2.26%. At discharge, these values improved to 51.58 ± 4.04 mm (LVIDd), 41.23 ± 5.30 mm (LVIDs), and 39.25% ± 3.75% (LVEF). One month postoperatively, further improvements were observed: 46.29 ± 3.76 mm (LVIDd), 37.45 ± 3.68 mm (LVIDs), and 43.22% ± 4.67% (LVEF). Group 2 showed similar positive outcomes, with preoperative values of 47.09 ± 5.06 mm (LVIDd), 35.11 ± 5.25 mm (LVIDs), and 50.13% ± 7.25% (LVEF), improving to 42.37 ± 4.18 mm (LVIDd), 31.05 ± 4.19 mm (LVIDs), and 55.33% ± 7.05% (LVEF) at one month postoperatively. Both groups demonstrated significant improvements in left ventricular function and NYHA class, with most patients moving from class III/IV to I/II. Complications were minimal, and no mortality was observed. Conclusion: OPCAB is safe and effective for patients with LVEF 30% - 39% and LVEF ≥ 40%, providing significant short-term functional improvements without increased risk.

Navigating Long-Term Management Challenges in Short Bowel Syndrome: A Case Report of Chronic Intestinal Failure Complicated by Kidney Dysfunction

The most common cause of intestinal failure is short bowel syndrome (SBS), occurring as a result of a small functional intestine length, usually less than 200 cm, leading to intestinal malabsorption. A 59-year-old female with a past medical history of Crohns disease status post total colectomy with ileostomy over 20 years ago came to the hospital due to progressive weakness. Despite medical management, the patient had high ileostomy output, leading to electrolyte disbalance, metabolic acidosis, dehydration, and progressive kidney decline. Due to the high dependence on continuous fluid supplementation, it was decided to place a port for parenteral hydration to maintain fluid replacements and homeostasis after discharge. Prompt initiation of parenteral fluid replacement and close follow-up on patients with ileostomy and intestinal failure is strongly recommended to avoid complications and prevent intestinal, liver, or kidney transplants.

The Relationship between Extreme Precipitation Events in East Africa during the Short Rainy Season and Indian Ocean Sea Surface Temperature

The East African short rainy season (October-November-December) is one of the major flood seasons in the East African region. The amount of rainfall during the short rainy season is closely related to the lives of the people and the socio-economic development of the area. By using precipitation data and sea surface temperature data, this study reveals the spatial and temporal variation patterns of extreme precipitation during the East African short rainy season. Key findings include significant rainfall variability, with Tanzania experiencing the highest amounts in December due to the southward shift of the Intertropical Convergence Zone (ITCZ), while other regions receive less than 100 mm. Extreme rainfall events (90th percentiles) are evenly distributed, averaging 2 to 10 days annually. Historical data shows maximum seasonal rainfall often peaks at 15 mm, with frequent occurrences of daily rainfall exceeding 10 mm during OND. Additionally, a positive correlation (0.48) between OND precipitation extremes and Indian Ocean Dipole (IOD) anomalies is statistically significant. These findings highlight the climatic variability and potential trends in extreme rainfall events in East Africa, providing valuable insights for regional climate adaptation strategies.

Saturnin-Short轻量级认证加密算法的统计无效故障分析

面向随机单字节故障模型和唯密文攻击假设,提出了一种针对Saturnin-Short算法的统计无效故障分析方法。该方法基于统计分布和无效状态分析,通过结合故障注入前后中间状态的变化,设计并采用了概率对称卡方-极大似然估计和调和中项-汉明重量新型区分器,最少仅需1097个无效故障并以不低于99%的成功率恢复Saturnin-Short算法的256 bit原始密钥。实验分析表明,所提区分器不仅降低了故障注入数,而且减少了攻击时间和复杂度。因此,Saturnin-Short算法不能抵抗统计无效故障分析的攻击。研究结果为其他轻量级认证加密算法的安全性分析提供了重要参考。

基于CNN-SAEDN-Res的短期电力负荷预测方法

基于深度学习的序列模型难以处理混有非时序因素的负荷数据,这导致预测精度不足。提出一种基于卷积神经网络(CNN)、自注意力编码解码网络(SAEDN)和残差优化(Res)的短期电力负荷预测方法。特征提取模块由二维卷积神经网络组成,用于挖掘数据间的局部相关性,获取高维特征。初始负荷预测模块由自注意力编码解码网络和前馈神经网络构成,利用自注意力机制对高维特征进行自注意力编码,获取数据间的全局相关性,从而模型能根据数据间的耦合关系保留混有非时序因素数据中的重要信息,通过解码模块进行自注意力解码,并利用前馈神经网络回归初始负荷。引入残差机制构建负荷优化模块,生成负荷残差,优化初始负荷。算例结果表明,所提方法在预测精度和预测稳定性方面具有优势。

基于阻抗在线测量的锂离子电池过放电诱发内短路识别研究

锂离子电池的过放电行为可诱发内短路进而可能导致热失控。由于单体电池一致性差异和过放电导致的内短路初期电、热特征不明显,使得电压和温度等常规物理参数难以可靠实现故障预警,而阻抗可以反映电池内部信息,对故障状态具有较好的指示能力。该文通过分析锂离子电池过放电诱发内短路引起阻抗变化的机理,基于设计的阻抗在线测量装置,确定用于监测电池内短路故障的特征阻抗频率为70 Hz,获取过放电诱发内短路过程中特征频率下动态阻抗及阻抗变化率,提出一种基于动态阻抗特征的内短路在线识别方法并验证了方法的可靠性。实验结果表明,锂离子电池放电过程中动态阻抗半正弦变化特征可提前约144 s实现过放电预警,动态阻抗针状变化特征可提前约152 s实现内短路故障预警,动态阻抗明显回升特征可作为内短路发生的标志。此外,阻抗变化率特征有助于实现锂离子电池故障识别及预警,该方法在锂离子电池故障在线快速诊断中具有重要的应用潜力。

旅游短视频营销对顾客融入的影响机制研究

以SIPS(sympathize-identify-participate-spread)模型为基础,对417名旅游短视频观看者进行问卷调查,通过实证研究构建旅游短视频营销对顾客融入的影响机制模型,旨在探索旅游目的地形象宣传的新方向和旅游短视频创作重点。研究发现:旅游短视频营销从传播内容、传播场景和传播主体3个方面正向影响顾客融入,顾客灵感具有部分中介作用;关系强度正向调节旅游短视频营销与顾客灵感的关系,旅游短视频营销通过顾客灵感正向影响顾客融入的间接效应,在高关系强度条件下更强烈。

基于CEEMD-SE的CNN&LSTM-GRU短期风电功率预测

为进一步提升短期风电功率的预测精度,提出了一种基于互补集合经验模态分解-样本熵(complementary ensemble empirical mode decomposition-sample entropy,CEEMD-SE)的卷积神经网络(convolutional neural network,CNN)和长短期记忆-门控循环单元(longshorttermmemory-gatedrecurrentunit,LSTM-GRU)的短期风电功率预测模型。首先,利用互补集合经验模态分解将原始风电功率序列分解为若干本征模态函数(intrinsic mode function,IMF)分量和一个残差(residual,RES)分量,利用样本熵算法将相近的分量进行重构;其次,搭建卷积神经网络和长短期记忆网络的并行网络结构,提取数据的局部特征和时序特征,并将特征融合后输入门控循环单元网络中进行学习预测;最后,通过算例进行验证,结果表明采用该模型后预测精度得到了有效提升,其均方根误差降低了15.06%、平均绝对误差降低了15.22%、决定系数提高了1.91%。

基于多维气象信息时空融合和MPA-VMD的短期电力负荷组合预测模型

为提高电力负荷预测精度,需考虑区域内不同地区多维气象信息对电力负荷影响的差异性。在空间维度上,提出多维气象信息时空融合的方法,利用Copula理论将多座气象站的风速、降雨量、温度、日照强度等气象信息与电力负荷进行非线性耦合分析并实现时空融合。在时间维度上,采用海洋捕食者算法(MPA)实现变分模态分解(VMD)核心参数的自动寻优,并采用加权排列熵构造MPA-VMD适应度函数,实现负荷序列的自适应分解。通过将时间维度各分量与空间维度各气象信息进行融合构造长短期记忆(LSTM)网络模型与海洋捕食者算法-最小二乘支持向量机(MPA-LSSVM)模型的输入集,得到各分量预测结果,根据评价指标选择各分量对应的预测模型,重构得到整体预测结果。算例分析结果表明,所提预测模型优于传统预测模型,有效提高了电力负荷预测精度。

助行康复机器人自适应步态控制方法研究

为了提高助行康复机器人在不同坡度上的稳定性,合理的步态控制方法显得尤为重要。这里以助行康复机器人为研究对象,提出了一种将长短期记忆网络和动态运动基元结合的步行康复机器人自适应步态控制方法。将传感器采集信息与动态运动基元中步态控制项相结合,对参考步态进行调整。同时通过更改动态运动基元的时间项和空间项对步态进行二次调整以适应不同的坡度。通过试验对该控制方法的可用性进行验证。结果表明,该方法可以在(0~20)°斜坡进行自适应步态变换,具有一定的实用价值。

基于残差修正CNN-BiLSTM的空中目标航迹短期预测算法

针对因深度学习自身局限性和递归预测策略产生的累积误差,导致航迹预测精度不高的问题,提出了一种基于残差修正CNN-BiLSTM的空中目标航迹短期预测算法。首先,引入卷积模块用于提取航迹数据之中具有潜在关联的空间位置特征,利用双向长短时记忆网络提取航迹数据中的时序特征,并实现对空中目标的实时单步预测和多步超前预测;其次,引入整合移动平均自回归为残差修正模型,对实时单步预测产生的残差建模,计算混合神经网络模型多步超前预测时的残差值;最后,将混合神经网络模型和残差修正模型的输出结果进行融合,得到最终的航迹预测值。实验结果表明,该算法大大降低了神经网络因自身局限性产生的误差和因递归策略预测产生的累积误差,能够显著提高空中目标航迹短期预测的精度。

基于改进麻雀搜索算法优化LSTM的滚动轴承故障诊断

为了对滚动轴承的工作状态及故障类别进行准确的诊断,本文采用长短时记忆(LSTM)神经网络作为分类器对滚动轴承数据集进行分类诊断。首先,从滚动轴承原始运行振动信号中提取时域和频域特征参数,组成具有高维特征参数的数据集;使用核主成分分析(KPCA)方法对高维特征集进行降维处理,选取重要性程度高的特征构成输入特征向量。然后,针对LSTM神经网络在滚动轴承故障诊断中存在的超参数难以确定的问题,提出一种基于自适应t分布策略的麻雀搜索算法优化LSTM神经网络的故障诊断方法(tSSA–LSTM)。最后,使用凯斯西储大学滚动轴承数据中心的数据进行故障诊断精度测试、泛化性能测试及噪声环境下故障诊断性能测试等多个仿真实验,并将本文提出的诊断模型与麻雀搜索算法优化长短时记忆神经网络(SSA–LSTM)、遗传算法优化长短时记忆神经网络(GA–LSTM)、粒子群算法优化长短时记忆神经网络(PSO–LSTM)及传统LSTM诊断模型进行对比。结果表明:tSSA可以更有效地对LSTM的隐含层神经元数量、周期次数、学习率等超参数进行合理优化;所提方法的平均诊断准确率达到98.86%,交叉验证平均诊断结果为98.57%;所提方法在噪声干扰下的故障诊断准确率也优于对比方法。因此,本文提出的tSSA–LSTM模型不仅可以更精准地诊断滚动轴承故障状态,而且具有更强的泛化能力及抗干扰能力,有效地提高了滚动轴承故障诊断的性能。

从视觉中心主义到审美资本主义——以抖音为例的短视频文化解读

以短视频为代表的图像文化已进入到人们生活的各个领域,图像无处不在。短视频成为我们生活中“物”的一种类型,视觉文化不只是沟通和反映我们的生活世界,也是在创造世界。从短视频对当今社会全方位的影响程度看,它充分体现了“视觉中心主义”或“视界政体”特征。短视频系统构建了“全景敞视视界”体系,以统计学的方法形成了“全景敞视监狱”式的监视模式,并以“投其所好”的推送方式实现对观看者的无形规训;人们的窥视本能与短视频的“审美诱惑”相结合,以审美的名义实现资本利益的获取,实为一种审美资本主义。

高速铁路0.01~1 m波段高低轨道不平顺特性分析

随着我国高速铁路发展规模的不断扩大及运营速度的不断提升,轨道短波不平顺引起的振动与噪声问题日益严重。深入研究高速铁路短波不平顺的时域、频域与时频域特征,对精准掌握高速铁路轨道短波平顺状态及其引起的车-轨系统振动、伤损等一系列问题具有重要的参考价值。以某高速铁路实测0.01~1 m波段轨面高低不平顺数据为研究对象,经过数据预处理剔除异常值并消除趋势项后,对实测不平顺数据进行时域分析,得到该高铁线路短波不平顺数据的统计与分布特性。采用平均周期图法计算该高铁线路轨道短波不平顺功率谱密度,应用非线性最小二乘拟合优化算法进行拟合,提出高速铁路轨道短波不平顺谱。从时频域角度对该段数据进行经验模态分解与希尔伯特变换,分析该高铁线路短波不平顺的希尔伯特谱与边际谱。研究结果表明:该高铁线路轨道短波不平顺数据各统计量幅值都较小,且近似服从正态分布;该高铁线路轨道短波不平顺功率谱密度存在与钢轨波磨有关的多个周期性成分,提出的高速铁路短波不平顺谱函数能够很好地对实测功率谱密度进行拟合,且拟合效果优于经典的SATO建议的短波谱与50 kg/m钢轨的短波谱;对该高铁线路轨道可能存在的短波病害进行识别与定位,时频域分析的方法可以较好地应用于高速铁路轨道平顺性的分析与日常的管理维护。

金融科技赋能下供应链金融对企业价值的影响

发展供应链金融对于深化金融供给侧结构性改革,增强金融服务实体经济具有不容忽视的重要战略意义.本文通过对2007年-2019年中国A股上市公司全部320.8万篇公告的文本数据信息进行收集整理,实证检验了上市公司发展供应链金融业务对企业价值的影响.研究结果表明:发展供应链金融在短期可以促进企业股价的上涨,在长期可以促进企业价值的提升,并且上市公司采用的供应链金融业务越多,公告中提到供应链金融词汇越频繁,这种提升效果越显著.发展供应链金融业务可以通过信号传递效应、风险承担效应、系统管理效应促进企业价值提高.在供应链金融相关公告中提到金融科技词汇的词频越高,上市公司获得的企业价值提升效果越强.金融科技可以为股票流动性较低、风险承担较弱的企业发展供应链金融业务起到更强的赋能效果.在进行一系列稳健性检验之后,上述结论依然成立.本文不仅为中国公司发展供应链金融业务提高企业价值,提供了基于公开市场数据大样本的实证证据,也对国家和地方政府支持供应链金融的发展具有现实启示。

基于BERT-BiLSTM-CRF模型的畜禽疫病文本分词研究

针对畜禽疫病文本语料匮乏、文本内包含大量疫病名称及短语等未登录词问题,提出了一种结合词典匹配的BERT-BiLSTM-CRF畜禽疫病文本分词模型。以羊疫病为研究对象,构建了常见疫病文本数据集,将其与通用语料PKU结合,利用BERT(Bidirectional encoder representation from transformers)预训练语言模型进行文本向量化表示;通过双向长短时记忆网络(Bidirectional long short-term memory network,BiLSTM)获取上下文语义特征;由条件随机场(Conditional random field,CRF)输出全局最优标签序列。基于此,在CRF层后加入畜禽疫病领域词典进行分词匹配修正,减少在分词过程中出现的疫病名称及短语等造成的歧义切分,进一步提高了分词准确率。实验结果表明,结合词典匹配的BERT-BiLSTM-CRF模型在羊常见疫病文本数据集上的F1值为96.38%,与jieba分词器、BiLSTM-Softmax模型、BiLSTM-CRF模型、未结合词典匹配的本文模型相比,分别提升11.01、10.62、8.3、0.72个百分点,验证了方法的有效性。与单一语料相比,通用语料PKU和羊常见疫病文本数据集结合的混合语料,能够同时对畜禽疫病专业术语及疫病文本中常用词进行准确切分,在通用语料及疫病文本数据集上F1值都达到95%以上,具有较好的模型泛化能力。该方法可用于畜禽疫病文本分词。

结合太阳辐射量计算与CNN-LSTM组合的光伏功率预测方法研究

为了提高模型预测性能,提出一种综合太阳辐射模型及深度学习的光伏功率预测模型。首先,利用太阳辐射机理建立太阳辐射模型(SRM),估算出水平面上总辐射值,再由斜面辐照度转换方法计算出光伏组件所接收的斜面辐射值。其次,通过皮尔逊相关分析法筛选出对光伏功率影响较大的主要因素,将斜面辐射计算值及主要影响因素作为输入,采用卷积神经网络(CNN)和长短期记忆网络(LSTM)建立光伏功率SRM-CNN-LSTM预测模型。分别利用春夏秋冬四季典型日的数据开展对比实验,结果表明:与几种其他方法相比,该文方法具有更好的预测效果。

Stress Distributions Created by Short and Regular Implants Placed in the Anterior Maxilla at Different Angles: A Finite Element Analysis

Although short implants are seen as alternative treatments that require additional surgical techniques in posterior region, they can be applied to anterior maxilla and various studies are required on this subject. The purpose of this study was to examine and compare the peak von Mises stress distributions in the crown, implant and abutment by using finite element analysis (FEA). Besides, a comparison of the implant-abutment connection types in the short implant with the FEA method was established. A short implant (4 × 5 mm) with a taper-lock connection and a regular implant (4 × 9 mm) with a screw connection were used in maxillary central incisor tooth area. Three different titanium abutments with 0?, 15? and 25? angles were used for abutments. In addition, in order to determine whether the stress change in short implants is due to the length of the implant-abutment connection, a screw was designed for a short implant and it was also evaluated in the same three angles. A total of three groups and nine models were generated. 114.6N load was applied to the cingulum area of the crown at an angle of 135? to the long axis of the crowns. A torque load of 25 Ncm was applied to the regular and short implant screw. Von Mises stress distributions of implants, abutments and crowns were evaluated by using FEA. Increased angle in implants increased von Mises stress values of implant, abutment and crown. Screw connection was found higher at all angles in short implants. Close values were found at different angles in taper-lock short implant crowns. The length and the angle in the bone of implant with the type of implant-abutment connection results in the accumulated stress values. Clinical Implications Taper implant-abutment connection system was found to be more promising in terms of stress accumulation in crowns. Although the amount of stress on the abutment increased due to the length of the implant in short implants, taper implant-abutment connection system slightly reduced related to this increase.