Fuzzy Cluster Analysis and Prediction of Psychiatric Health Data Based on BPNN

PMH (psychiatry/mental health) is affected by many factors, among which there are numerous connections, so the prediction of PMH is a nonlinear problem. In this paper, BPNN (Back Propagation Neural Network) is applied to fuzzy clustering analysis and prediction of PMH data, and the rules and characteristics of PMH and behavioral characteristics of people with mental disorders are analyzed, and various internal relations among psychological test data are mined, thus providing scientific basis for establishing and perfecting early prevention and intervention of mental disorders in colleges and universities. Artificial neural network is a mathematical model of information processing, which is composed of synapses similar to the structure of brain neurons. The fuzzy clustering analysis and data prediction ability of optimized PMH data are obviously improved. Applying BPNN to the fuzzy clustering analysis and prediction of PMH data, analyzing the rules and characteristics of PMH and the behavioral characteristics of patients with mental disorders, can explore various internal relations among psychological test data, and provide scientific basis for establishing early prevention and intervention of mental disorders.

Running Status Diagnosis of S700K Turnout Based on VMD-KPCA and Fuzzy Clustering

S700K turnout is the key equipment of railway line conversion. The diagnosis of S700K turnout in a normal, sub-health, and fault running state is the primary premise to ensure the safe operation of the railway. Aiming at the consistency between the characteristics of the power curve of S700K turnout and its state information, this paper proposes a new algorithm based on variational mode decomposition (VMD) and kernel principal component analysis (KPCA) to extract the characteristics of the power curve of S700K turnout. It uses fuzzy clustering analysis to diagnose the running state of S700K turnout. First, to extract the detailed components of the action power curve, it is decomposed into intrinsic mode function with limited bandwidth (BIMF) by VMD. Secondly, the multi-scale permutation entropy (MPE) is used to characterize the signal complexity of the power curve and different BIMF components, which are taken as the running state feature set. After KPCA analysis, eigenvalues with a contribution rate greater than 95% are selected as the state eigenvector. The experimental results show that the diagnosis algorithm can effectively identify the running state of S700K turnout, meet the characteristics of fewer fault samples of S700K turnout, and do not need to train in advance, which is of great significance for field guidance.

Importance of Regional Rainfall Data in Homogeneous Clustering of Data Sparse Areas: A Study in the Upper Brahmaputra Valley Region

Delineation of homogeneous regions has found its way into many hydrological applications as it helps in addressing the challenges in understanding the behavior of rainfall distribution and its variability at a local scale. In the present study, rainfall data recoded by 83 tea gardens in the upper Brahmaputra valley region of Assam have been used to identify homogeneous rainfall regions by using fuzzy clustering analysis. Further, seven different cluster validity indices (CVs) were utilized to find out the optimum clustering in the fuzzy c-means (FCM) algorithm. The clusters thus formed were assessed for statistical homogeneity by performing homogeneity tests based on L-moment. Three different combinations of feature vectors were employed in FCM algorithm and the outputs were compared for attaining best solutions to regionalization. The results were further compared with previous regionalization studies. The analysis and comparison conclude that if regionalization needs to be done at a local scale, further sub-clustering of a larger clustered region to smaller regions may be required. Local rainfall data can be used for the purpose provided a good dataset with large number of station points are available within the region. Along with rainfall data, geographical location parameters (latitude, longitude and elevation) need to be taken into account for getting a definite conclusion.

Fuzzy clustering analysis of comprehensive hand of polyester fabric based on the CHES-FY system

Polyester fabrics are attributed with various performances and are currently applied widely in textiles. This necessitates a quick and effective selection process to choose polyester fabrics to correspond with engineers' designs for industrial textiles. Therefore, the main focus of this paper is to present the comprehensive handle evaluation system for fabrics and yarns (CHES-FY), which has been specially developed to measure the basic handle of textile materials, including softness, stiffness, smoothness and tightness. Several kinds of polyester fabrics were chosen to undergo subjective evaluation and testing by the CHES-FY system, and were assigned into corresponding clusters by the K-means cluster method. The basic hand indexes of polyester fabric were featured. Comparisons between subjective judgments and the objective K-means cluster method were conducted. Experimental results show that a good correlation exists between subjective judgment and the objective cluster method, indicating that the four basic hand indexes measured by the CHES-FY system can be utilized to characterize the comprehensive hand of industrial polyester fabrics, and that the CHES-FY system can be used to discriminate categories of polyester fabrics.

Entropy Weighting Based Centralized Multi-View Fuzzy Clustering: A Case Study on Brain MR Image Segmentation

Multi-view fuzzy clustering analysis is often used for medical image segmentation such as brain MR image segmentation. However, in traditional multi-view clustering, it assumes that each view plays the same role to the final partition result, which omits the negative influences caused by noisy or weak views. In this paper, a novel entropy weighting based centralized clustering technique is proposed for multi-view datasets where the Shannon entropy is hired for view weight learning. Moreover, the centralized strategy is employed for collaborate learning. Extensive experiments show that the promising performance of our proposed clustering technique. More importantly, a case study on brain MR image segmentation indicates the application ability of our clustering technique.

Prevalence of Acute Myocardial Infarction and Changing Meteorological Conditions in Iran: Fuzzy Clustering Approach

Background: The prevalence of Acute Myocardial Infarction (AMI) varies from region to region caused by seasonal climate changes and temperature variation. This study aimed to assess the relationship between changing meteorological conditions and incidence of AMI in Iran. Methods: This retrospective prevalence study was based on medical records of the heart center of Mazandaran Province on all patients diagnosed with AMI in Mazandaran, northern Iran between 2013 and 2015. Patients’ sex and the day, month, year and time of hospital admission were extracted from patients’ records. Moreover, the meteorological reports were gathered. Results: A statistically significant difference was found between the distributions of AMI cases across 12 months of the year (P < 0.01). Fuzzy clustering analysis using 16 different climatic variables showed that March, April, and May were in the same cluster together. The other 9 months were in different clusters. Conclusion: Significant increase in AMI was seen in March, April and May (cold to hot weather).