基于改进粒子群优化和极限学习机的网络安全态势预测

针对网络安全态势预测模型预测精度不高、收敛较慢等问题,提出了一种基于改进粒子群优化极限学习机（IPSO-ELM）算法的预测方法。首先,通过改进粒子群优化（PSO）算法中的惯性权重和学习因子来实现两种参数随着迭代次数增加的自适应调整,使PSO初期搜索范围大、速度高,后期收敛能力强、稳定。其次,针对PSO易陷入局部最优的问题,提出一种粒子停滞扰动策略,将陷入局部最优的粒子重新引导至全局最优飞行。改进粒子群优化（IPSO）算法既保证了全局寻优的能力,又对局部搜索能力有所增强。最后,将IPSO与极限学习机（ELM）结合来优化ELM的初始权值及阈值。与ELM相比,结合IPSO的ELM的预测精度提高了44.25%。实验结果表明,与PSO-ELM相比,IPSO-ELM的预测结果拟合度可达到0.99,收敛速度提升了47.43%。所提算法在预测精度和收敛速度等指标上明显优于对比算法。     

无线传感器网络中基于ELM的混合入侵检测方案

在研究机器学习算法的基础上,提出一种基于极限学习机( ELM)的混合入侵检测方案。将无线传感器网络分为感知层、数据汇聚层和核心控制层,在每层分别设置与其相适应的入侵检测方案,并在能量充足的核心控制层布置信任管理模块和ELM模块。信任模块可以及时筛去异常节点,相比于支持向量机算法训练速度更快,可提高入侵检测效率。实验结果表明,该方案在保证较高检测率的基础上,降低了能耗,延长网络运行时间,更适合于资源受限的无线传感器网络。     

结合栈式监督AE与可变加权ELM的回归预测模型

在现代工业生产过程中,许多关键变量与产品质量或生产效率密切相关,关键变量的实时监测是实现利润最大化及节能降耗的有效途径。针对回归预测任务中目标特征提取不全面、预测精度较低等问题,提出一种基于栈式监督自编码器与可变加权极限学习机的回归预测模型。通过堆叠多层自编码器并在每层自编码器中添加回归网络,同时以有监督方式对栈式自编码器（SAE）进行逐层预训练,得到与输出变量相关的特征表示。利用反向传播算法对网络参数进行微调,优化自编码器模型参数。在分析提取特征与输出变量的相关性基础上,对极限学习机（ELM）的输入权值和偏置进行加权得到预测结果。实验结果表明,与基于ELM和SAE-ELM的回归预测模型相比,该模型在多晶硅铸锭的G6产品数据集上的均方根误差降低0.056 7和0.011 2、决定系数提高0.489 3和0.290 3,具有更高的回归预测准确性及更强的鲁棒性与泛化性能。     

The moderating effects of psychological reactance and product involvement on online shopping recommendation mechanisms based on a causal map

The use of the Internet in the daily activities of individuals and firms has become entrenched, and online shopping has become commonplace. However, debates about how online shopping recommendation mechanisms (OREMs) should be designed have not been completely resolved. The challenge with traditional online shopping recommendation mechanisms (TR-OREMs) is that they focus too much on quantitative factors. Thus, they ignore causal interrelationships with qualitative factors that are believed to significantly impact quantitative factors. Considering only quantitative factors and ignoring qualitative ones likely distorts the final recommendation results. Another problem with TR-OREMs is that they ignore the perceived psychological reactance of consumers against the recommended products. Such consumer reactance may be reduced when the causal interrelationships among all the quantitative and qualitative factors are analyzed and incorporated properly into the OREM. To overcome these problems, we propose a causal map – online shopping recommendation mechanisms (CM-OREMs) based on a causal map. We analyzed possible causal relationships among quantitative and qualitative factors and incorporated them in the recommendation process to reduce consumer reactance against the recommendation results. Furthermore, an elaboration likelihood model (ELM) was used to build hypotheses about how the online shopping behavior of consumers is affected by OREMs based on the proposed causal map. Specifically, the performance of the proposed OREM was empirically analyzed by gathering experiment data from qualified respondents who were asked to refer to the proposed OREM before making purchasing decisions via mobile phones. Statistical results showed that the proposed OREMs could enhance consumer decision satisfaction, decision confidence, and attitude toward the recommended products. It could also positively affect consumer purchasing intentions. The OREM had a greater effect on the high-reactance group of participants than on the low-reactance group as well as on a high-involvement product versus a low-involvement product.     

Prediction of landslide displacement with controlling factors using extreme learning adaptive neuro-fuzzy inference system (ELANFIS)

Landslide is a major geo-environmental hazard which imparts serious threat to lives and properties. The slope failures are due to adverse inherent geological conditions triggered by an external factor. This paper proposes a new method for the prediction of displacement of step-like landslides, by accounting the controlling factors, using recently proposed extreme learning adaptive neuro-fuzzy inference system (ELANFIS) with empirical mode decomposition (EMD) technique. ELANFIS reduces the computational complexity of conventional ANFIS by incorporating the theoretical idea of extreme learning machines (ELM). The rainfall data and reservoir level elevation data are also integrated into the study. The nonlinear original landslide displacement series, rainfall data, and reservoir level elevation data are first converted into a limited number of intrinsic mode functions (IMF) and one residue. Then decomposed displacement data are predicted by using appropriate ELANFIS model. Final prediction is obtained by the summation of outputs of all ELANFIS sub models. The performance of proposed the technique is tested for the prediction Baishuihe and Shiliushubao landslides. The results show that ELANFIS with EMD model outperforms other methods in terms of generalization performance.     

An elaboration likelihood model based longitudinal analysis of attitude change during the process of IT acceptance via education program

The principal objective of this study was to gain insight into attitude changes occurring during IT acceptance from the perspective of elaboration likelihood model (ELM). In particular, the primary target of this study was the process of IT acceptance through an education program. Although the Internet and computers are now quite ubiquitous, and thus many people can come to an acceptance of IT organically – including acceptance of such products as new released online services or software, some software – most notably programming languages and spreadsheets – may require education programs, and the efficacy of that education may determine the rate of acceptance. Thus, education programs for such IT may be regarded as not only teaching and training how to use the product from a technical standpoint, but also as a process by which program participants are persuaded to have a positive attitude towards accepting the new IT. This study adopted an ELM based longitudinal approach to capture the changes in the roles of persuasion routes and attitude during IT acceptance – in this case, the education program. To validate empirically the suggested model, junior students majoring in business administration, all of whom were taking an Excel class, were surveyed twice – just after mid-term and just after their final exam. These data were analysed via partial least-square method to deduce some possible implications. Our results supported all the hypotheses put forward.     

改进PSO优化ELM预测锂离子电池剩余寿命

针对极限学习机在预测锂离子电池剩余寿命过程中的不稳定性,提出利用混合粒子群优化算法对极限学习机预测模型优化的方法。通过改进的粒子群优化算法对极限学习机的输入端进行寻优处理,不但能够使模型的预测精度有进一步提高,而且大大增加了锂离子电池单次剩余寿命预测结果的可信度。利用NASA PCoE公开的锂离子电池数据进行仿真实验并评估该模型的预测性能,然后与标准的极限学习机预测模型预测结果进行对比,统计结果表明该方法使预测误差控制在2%左右。     

Critical Clearing Time prediction within various loads for transient stability assessment by means of the Extreme Learning Machine method

The Critical Clearing Time (CCT) is a key issue for Transient Stability Assessment (TSA) in electrical power system operation, security, and maintenance. However, there are some difficulties in obtaining the CCT, which include the accuracy, fast computation, and robustness for TSA online. Therefore, obtaining the CCT is still an interesting topic for investigation. This paper proposes a new technique for obtaining CCT based on numerical calculations and artificial intelligence techniques. First, the CCT is calculated by the critical trajectory method based on critical generation. Second, the CCT is learned by Extreme Learning Machine (ELM). This proposed method has the ability to obtain the CCT with load changes, different fault occurrences, accuracy, and fast computation, and considering the controller. This proposed method is tested by the IEEE 3-machine 9-bus system and Java-Bali 500 kV 54-machine 25-bus system. The proposed method can provide accurate CCTs with an average error of 0.33% for the Neural Network (NN) method and an average error of 0.06% for the ELM method. The simulation result also shows that this method is a robust algorithm that can address several load changes and different locations of faults occurring. There are 29 load changes used to obtain the CCT, with 20 load changes included for the training process and 9 load changes not included.     

基于改进的GA-ELM煤与瓦斯突出预测模型

为准确、快速地预测煤与瓦斯突出,提出了一种基于因子分析和遗传算法(GA)优化极限学习机(ELM)的煤与瓦斯突出危险性预测模型。构建10种影响因素的煤与瓦斯突出评价指标体系,采用因子分析法对评价指标体系进行分析提取,将提取出的5个公因子作为ELM的输入参数,为避免ELM输入权值和隐含层偏差随机性的影响,应用GA对ELM模型参数进行优化,构建GA-ELM模型,选取20组实例进行仿真预测,同时与传统单一的ELM、SVM和BP模型进行对比分析。结果表明:基于改进的GA-ELM模型能有效降低数据冗余、简化网络结构和提高判别精度,提出将其运用到煤与瓦斯突出的预测与实际结果具有很好的一致性。