Knowledge Extraction Algorithm for Variances Handling of CP Using Integrated Hybrid Genetic Double Multi-group Cooperative PSO and DPSO

Although the clinical pathway (CP) predefines predictable standardized care process for a particular diagnosis or procedure, many variances may still unavoidably occur. Some key index parameters have strong relationship with variances handling measures of CP. In real world, these problems are highly nonlinear in nature so that it’s hard to develop a comprehensive mathematic model. In this paper, a rule extraction approach based on combing hybrid genetic double multi-group cooperative particle swarm optimization algorithm (PSO) and discrete PSO algorithm (named HGDMCPSO/DPSO) is developed to discovery the previously unknown and potentially complicated nonlinear relationship between key parameters and variances handling measures of CP. Then these extracted rules can provide abnormal variances handling warning for medical professionals. Three numerical experiments on Iris of UCI data sets, Wisconsin breast cancer data sets and CP variances data sets of osteosarcoma preoperative chemotherapy are used to validate the proposed method. When compared with the previous researches, the proposed rule extraction algorithm can obtain the high prediction accuracy, less computing time, more stability and easily comprehended by users, thus it is an effective knowledge extraction tool for CP variances handling.     

自适应DNA免疫算法在化工软测量中的应用

将T-S模糊模型与RBF神经网络相结合,构成T-S模糊RBF神经网络,提出了一种自适应DNA免疫算法优化设计T-S模糊RBF神经网络的规则后件参数的方法。该方法采用基于抗体浓度和克隆选择的更新策略调节机制,能有效地保持抗体的多样性,避免早熟收敛。将该方法应用于延迟焦化汽油干点的软测量建模,仿真结果表明了DNA免疫遗传算法在T-S模糊神经网络系统优化设计中的有效性,并可获得较高精度的模型。     

一类参数不确定T-S模糊系统的鲁棒控制及仿真

针对一类具有参数不确定的T akag i-Sugeno(T-S)模糊系统,基于模糊区域的概念研究了其鲁棒控制问题。通过将不确定T-S模糊模型转换为不确定T-S模糊区域模型,并利用Lya-punov稳定性理论,导出了线性矩阵不等式(LM I)形式的鲁棒控制器设计方法。相对于传统设计方法,降低了采用线性矩阵不等式方法求解的难度,并具有良好的鲁棒性能。仿真结果验证了该方法的有效性。     

一种基于T-S模糊模型的自适应建模方法及其应用

提出了一种改进的基于T-S模糊RBF神经网络模型的辨识算法和自适应方法，采用模糊C均值聚类(FCM)算法划分输入输出数据空间，最后将该算法应用于丙烯腈收率的预报，仿真结果表明了这种基于T-S模糊模型的自适应建模方法的有效性。     

细乳液法制备聚硅氧烷-Ag纳米复合微球及其抗菌性

模糊建模是一种有效的非线性系统建模方法,因为非线性系统的复杂性,仍有很多问题难以处理。针对T-S模糊模型,提出了一种改进的建模及优化方法。首先,将快速搜索密度峰聚类和模糊C均值聚类(FCM)算法相结合,使用快速搜索密度峰聚类算法找到聚类个数和初始聚类中心后,再用FCM算法进行聚类;然后,通过最小二乘法辨识结论参数得到初始T-S模糊模型,使用改进的差分进化(DE)算法整体优化模型的结构和参数,获得最终的T-S模型;最后,选择代表性实例,使用MATLAB程序进行仿真分析和比较,验证了本文方法能有效提高T-S模糊模型的辨识精度和速度。     

基于粗糙集的T-S模糊神经网络在回转窑烧结过程中的应用

基于粗糙集理论的知识约简方法和T-S模糊神经网络的非线性映射理论,针对回转窑烧结过程被控对象复杂、各参数之间相互耦合及难以建立精确数学模型的特点,提出一种RS-FNN智能控制策略。采用基于一种新的聚类有效性准则函数的模糊C均值聚类算法对连续属性进行离散化;然后利用粗糙集理论由历史数据样本提取约简规则集,对应的T-S模型具有反映数据特征的良好拓扑结构;最后T-S模型参数由梯度下降混合最小二乘法进行精调。该方法应用于铁矿氧化球团回转窑生产过程控制取得了良好效果,增强了系统容错及抗干扰的能力。     

改进的PSO算法在重油热解模型参数估计中的应用

针对传统的粒子群算法(PSO)在解决复杂的优化问题时易陷入局部最优这一情况,提出了一种改进的粒子群算法(EPSO),该算法在传统的粒子群算法陷入局部最优的情况下引入了单个粒子的"Hooke-Jeeves模式搜索"操作和粒子之间的"启发式交叉"操作。仿真结果表明:EPSO算法的全局搜索性能和收敛速度比传统的PSO算法有明显的提高。采用EPSO算法进行非线性参数估计所得到的重油热解模型,其预报的平均相对误差比传统的PSO算法得到的模型提高了11.98%,比遗传算法(GA)得到的模型提高了38.76%。     

Neural Network Classifier with Entropy Based Feature Selection on Breast Cancer Diagnosis

The aim of this research is to combine the feature selection (FS) and optimization algorithms as the optimal tool to improve the learning performance like predictive accuracy of the Wisconsin Breast Cancer Dataset classification. An ensemble of the reduced data patterns based on FS was used to train a neural network (NN) using the Levenberg–Marquardt (LM) and the Particle Swarm Optimization (PSO) algorithms to devise the appropriate NN training weighting parameters, and then construct an effective Neural Network classifier to improve the Wisconsin Breast Cancers’ classification accuracy and efficiency. Experimental results show that the accuracy and AROC improved emphatically, and the best performance in accuracy and AROC are 98.83% and 0.9971, respectively.     

基于GAs/PSO组合算法的水轮机调速系统PID参数寻优

提出了一种基于GA s/PSO组合算法的P ID控制器参数自整定方法,这种方法兼有遗传算法(GA s)和粒子群算法(PSO)的优点。组合算法种群由GA s和PSO的最佳个体迁移形成,其中GA s采用了实数编码和变异概率自适应,PSO算法采用了带指数衰减的惯性因子的速度更新算法,以加快收敛速度。通过对水轮机调速系统P ID控制器参数寻优仿真比较表明,该组合算法寻优性能比单独的GA s和PSO表现更为优异,且所得系统具有更好的动态性能。     

Computer-Aided Intelligent System for Diagnosing Pediatric Asthma

Asthma is a lung chronic inflammatory disorder estimated between 1.4% and 27.1% in different area of the world. Result of various studies show that asthma is usually underdiagnosed especially in developing countries, because of limitations on access to medical specialists and laboratory facilities. In this paper, we report on the development and evaluation of a novel patient-based fuzzy system that promotes the diagnosis method of asthma. The design of this application addresses five critical issues included: 1) modular representation of asthma diagnostic variables regard to patients’ perception of the disease, 2) algorithmic approaches conducting inference of diagnosing based on patient’s response to questions, 4) front-end mechanism for capturing data from patient, 5) output for both patient and physician regard to asthma possibility. for the system output score (0–10) the efficacy of this system calculated in the study sample included 139 asthmatic patients and 139 non-asthmatic patients (age range 6–18) reinforce the sensitivity of 88% and specificity of 100% for cut off value 0.7.     

基于混合知识的自适应粒子群算法及其在博弈问题中的应用

粒子群优化算法的寻优性能往往会受到控制参数和速度策略的影响。为提高粒子群优化算法的性能,提出了一种基于混合知识的自适应粒子群算法（SPSO-MK）。该算法使用不同的速度更新策略来平衡粒子群优化算法的局部和全局搜索能力,利用在线和先验知识分别对惯性权重和加速因子进行调整。选取32个测试函数进行仿真实验,结果表明本文算法的整体性能好于10种粒子群的变种算法和3种非粒子群算法。将本文算法用于求解3个非合作博弈纳什均衡问题,结果表明该算法能够取得较好的结果。     

非满载车辆路径问题的改进粒子群优化算法

将局部版粒子群算法应用于非满载车辆路径问题,设计了一种实数编码方案,线性调整惯性权值,改进粒子更新公式,建立了解决该问题的粒子群算法。用该算法求解了两个车辆路径问题的算例,并与遗传算法和标准粒子群算法进行了比较。结果表明:该算法提高了搜索最优路径的成功率,能更有效地求解非满载车辆路径问题。     

A Hybrid Data Mining Model to Predict Coronary Artery Disease Cases Using Non-Invasive Clinical Data

Coronary artery disease (CAD) is caused by atherosclerosis in coronary arteries and results in cardiac arrest and heart attack. For diagnosis of CAD, angiography is used which is a costly time consuming and highly technical invasive method. Researchers are, therefore, prompted for alternative methods such as machine learning algorithms that could use noninvasive clinical data for the disease diagnosis and assessing its severity. In this study, we present a novel hybrid method for CAD diagnosis, including risk factor identification using correlation based feature subset (CFS) selection with particle swam optimization (PSO) search method and K-means clustering algorithms. Supervised learning algorithms such as multi-layer perceptron (MLP), multinomial logistic regression (MLR), fuzzy unordered rule induction algorithm (FURIA) and C4.5 are then used to model CAD cases. We tested this approach on clinical data consisting of 26 features and 335 instances collected at the Department of Cardiology, Indira Gandhi Medical College, Shimla, India. MLR achieves highest prediction accuracy of 88.4 %.We tested this approach on benchmarked Cleaveland heart disease data as well. In this case also, MLR, outperforms other techniques. Proposed hybridized model improves the accuracy of classification algorithms from 8.3 % to 11.4 % for the Cleaveland data. The proposed method is, therefore, a promising tool for identification of CAD patients with improved prediction accuracy.     

Classification of Benign and Malignant Breast Masses Based on Shape and Texture Features in Sonography Images

The purpose of this research was evaluating novel shape and texture feature’ efficiency in classification of benign and malignant breast masses in sonography images. First, mass regions were extracted from the region of interest (ROI) sub-image by implementing a new hybrid segmentation approach based on level set algorithms. Then two left and right side areas of the masses are elicited. After that, six features (Eccentricity_feature, Solidity_feature, DeferenceArea_Hull_Rectangular, DeferenceArea_Mass_Rectangular, Cross-correlation-left and Cross-correlation-right) based on shape, texture and region characteristics of the masses were extracted for further classification. Finally a support vector machine (SVM) classifier was utilized to classify breast masses. The leave-one-case-out protocol was utilized on a database of eighty pathologically-proven breast sonographic images of patients (forty-seven benign cases and thirty-three malignant cases) to evaluate our method. The classification results showed an overall accuracy of 95.00%, sensitivity of 90.91%, specificity of 97.87%, positive predictive value of 96.77%, negative predictive value of 93.88%, and Matthew’s correlation coefficient of 89.71%. The experimental results declare that our proposed method is actually a beneficial tool for the diagnosis of the breast cancer and can provide a second opinion for a physician’s decision or can be used for the medicine training especially when coupled with other modalities.     

基于改进云粒子群优化的模糊神经网络在甲醇合成转化率软测量中的应用

针对基本PSO算法早熟、搜索精度不高与易陷入局部最优的缺点,结合云滴的随机性、稳定倾向性,提出了一种改进粒子群优化算法(ICPSO)。将改进算法用于模糊神经网络的参数优化,并应用于甲醇单程转化率建模中。仿真实验结果表明：该模型具有较高的精度和较好的泛化能力,能够实现甲醇转化率的实时监测。     

A Biomedical Decision Support System Using LS-SVM Classifier with an Efficient and New Parameter Regularization Procedure for Diagnosis of Heart Valve Diseases

Classification success of Support Vector Machine (SVM) depends on the characteristic of given data set and some training parameters (C and σ). In literature, a few studies have been presented for regularization of these parameters which affects classification performance directly. This study proposes a new approach based on Renyi’s entropy and Logistic regression methods for parameter regularization. Our regularization procedure runs at two steps. In the first step, optimal value of kernel parameter interval is found via Renyi’s entropy method and optimal C value is found via logistic regression using exponential function in the next step. In addition to, this new decision support system is applied to biomedical research area via an application related to Doppler Heart Sounds (DHS). Experimental results show the efficiency of developed regularization procedure.     

Improved Particle Swarm Optimization Algorithm for Android Medical Care IOT using Modified Parameters

This study examines wireless sensor network with real-time remote identification using the Android study of things (HCIOT) platform in community healthcare. An improved particle swarm optimization (PSO) method is proposed to efficiently enhance physiological multi-sensors data fusion measurement precision in the Internet of Things (IOT) system. Improved PSO (IPSO) includes: inertia weight factor design, shrinkage factor adjustment to allow improved PSO algorithm data fusion performance. The Android platform is employed to build multi-physiological signal processing and timely medical care of things analysis. Wireless sensor network signal transmission and Internet links allow community or family members to have timely medical care network services.     

基于变精度粗糙-模糊集模型的诊断知识获取

提出一种基于变精度粗糙-模糊集模型的诊断知识获取算法,利用相似性聚类方法自动获取模糊隶属函数,将连续属性表示成模糊值,通过定义模糊相似关系和模糊相似类给出了变精度粗糙-模糊模型的近似表示,并引入蚁群算法求取模糊相似关系下的属性约简,进行诊断知识的获取。将其应用于精对苯二甲酸生产过程尾氧浓度故障诊断知识获取中,结果表明:该算法可以从故障数据中提取更客观有效的诊断规则,在实际故障诊断中具有很好的应用价值。     

一种基于高斯混合模型的快速水平集图像分割方法

水平集方法(LSM)图像分割的本质是求解一个随时间变化的偏微分方程,而使用变分法求解此水平集方程(LSE)往往要耗费过多的计算时间。为了减少算法的运行时间,提出了一种快速水平集图像分割算法。该算法在模糊聚类水平集方法(FCM-LSM)的基础上使用高斯混合模型(GMM)改造其隶属度损失函数,并利用离散网格Boltzmann方法(LBM)求解水平集方程。实验结果表明:本文提出的算法无论是在执行效率上还是在分割效果上都优于传统方法,证明了算法的可行性。     

Parallel Algorithm to Analyze the Brain Signals: Application on Epileptic Spikes

In the current work, we have proposed a parallel algorithm for the recognition of Epileptic Spikes (ES) in EEG. The automated systems are used in biomedical field to help the doctors and pathologist by producing the result of an inspection in real time. Generally, the biomedical signal data to be processed are very large in size. A uniprocessor computer is having its own limitation regarding its speed. So the fastest available computer with latest configuration also may not produce results in real time for the immense computation. Parallel computing can be proved as a useful tool for processing the huge data with higher speed. In the proposed algorithm ‘Data Parallelism’ has been applied where multiple processors perform the same operation on different part of the data to produce fast result. All the processors are interconnected with each other by an interconnection network. The complexity of the algorithm was analyzed as Θ((n + δn) / N) where, ‘n’ is the length of the input data, ‘N’ is the number of processor used in the algorithm and ‘δn’ is the amount of overlapped data between two consecutive intermediate processors (IPs). This algorithm is scalable as the level of parallelism increase linearly with the increase in number of processors. The algorithm has been implemented in Message Passing Interface (MPI). It was tested with 60 min recorded EEG signal data files. The recognition rate of ES on an average was 95.68%.