Self-organizing Map Network Research Articles

Maximum correntropy self-organizing map for waveform classification

The self-organizing map (SOM) is a well-known unsupervised classification method for seismic facies analysis. The conventional SOM networks and their mutation methods are mostly based on the Euclidean distance measurement, which estimates the global similarity between the input data and weights of neuron. Such measurement is suitable for signals with a Gaussian distribution but may not work well for seismic data for which the dominant structural features (e.g., stratigraphic layers, faults, channels, and salt bodies) typically appear some specially distributed patterns. Therefore, we propose to modify the SOM by introducing a new distance measurement criterion based on the maximum correntropy. The correntropy is mainly used to measure the local similarity between variables, especially for non-Gaussian signals, and therefore is more suitable than the Euclidean distance to adaptively characterize the seismic structural similarities for seismic classification tasks. Validations on two synthetic data sets indicate that the proposed method yields more accurate classification results and is more robust to noise than the traditional SOM algorithm. Comparison experiments on a field data example also demonstrate that the proposed method can delineate the distribution of seismic facies more effectively and stably than the traditional one.

Industrial users load pattern extraction method based on multidimensional electrical consumption feature construction

The rapid development of renewable energy generation aggravates the imbalance between supply and demand in power grid, and exploring the potential of demand side resource can effectively improve such problems. Industrial users (IU) is an important demand response resource of power grid, and mining the load patterns of IU is the basis of studying the demand response ability of IU, which plays an important role in the safe operation and lean management of power grid. Lately, the popularity of advanced metering infrastructures provides data support for studying the load patterns of IU. However, the high dimensionality and the complex non-linear relationship of IU’s load data bring difficulties to the task of clustering. To solve the above problems, this paper proposes a load pattern extraction method based on multidimensional electrical consumption feature construction. Firstly, industrial load characteristic set of IU is created with five load characteristic indexes weighted by improved entropy weight method. In addition, convolutional autoencoder is established to extract the temporal feature of industrial load data which is combined with industrial load characteristic set to build a multidimensional feature set (MFS) for IU and finish multidimensional electrical consumption feature construction (MECFC). Then, MFS is used as the input of Self-Organization Map network to select the initial clustering centers of K-means algorithm, overcoming the problem of local optimal solution, and complete the IU daily load clustering. The experiment shows that the algorithm based on MECFC solves the local optimal problem and have better performance in stability and clustering effect than traditional methods.

A Sorting Method of SAR Emitter Signal Sorting Based on Self-Supervised Clustering

Most existing methods for sorting synthetic aperture radar (SAR) emitter signals rely on either unsupervised clustering or supervised classification methods. However, unsupervised clustering can consume a significant amount of computational and storage space and is sensitive to the setting of hyperparameters, while supervised classification requires a considerable number of labeled samples. To address these limitations, we propose a self-supervised clustering-based method for sorting SAR radiation source signals. The method uses a constructed affinity propagation-convolutional neural network (AP-CNN) to perform self-supervised clustering of a large number of unlabeled signal time-frequency images into multiple clusters in the first stage. Subsequently, it uses a self-organizing map (SOM) network combined with inter-pulse parameters for further sorting in the second stage. The simulation results demonstrate that the proposed method outperforms other depth models and conventional methods in the environment where Gaussian white noise affects the signal. The experiments conducted using measured data also show the superiority of the proposed method in this paper.

Discovering risk patterns in people with affective disorder-induced disabilities associated with their healthcare delay.

People with affective disorder-induced disabilities (ADIDs) often experience complex needs that delay their healthcare. Discovering hidden patterns in these people for real-world use of health services is essential to improve healthcare delivery. A cross-sectional study population (2501 adults with ADIDs) was obtained from the Australian national representative survey of disability in 2015, including 21 demographic, health and social characteristics and healthcare delay information in general practice, specialist and hospital services. The Self-Organising Map Network was used to identify hidden risk patterns associated with healthcare delay and investigate potential predictors of class memberships by means of simple visualisations. While experiencing disability avoidance showed across different healthcare delays, labour force appeared not to have any influence. Approximately 30% delayed their healthcare to general practice services; these were young, single females in great need of psychosocial support and aids for personal activities. Those who delayed their healthcare commonly presented a lack of social connections and a need for contact with family or friends not living in the same household. The pattern evidence provides an avenue to further develop integrated care strategies with better targeting of people with ADIDs, considering social participation challenges facing them, to improve health service utilisation.

Examining global warming factors using self-organizing map and Granger causality network: a case from South Korea

BackgroundUnderstanding and patterning the possible causal variables of global warming is attributed to the development of effective prevention and mitigation strategies for climate change. Therefore, we aimed to pattern and visualize the possible causal variables of global warming and measure the causality between them.MethodsWe patterned and visualized the time series (103 years, from 1918 to 2020) of global surface temperature (GTemp) data with the gross domestic product (GDP) per capita, human population (Pop), and carbon dioxide (CO2) emissions of South Korea using a self-organizing map (SOM) and examined the causable local feature of global warming using the Granger causality (GC) test. The time-series data were trained and mapped in 4 × 4 SOM grids, and causality networks between variables were examined using multivariate Granger test statistics.ResultsSOM patterned 103 years of data, and a dominant cluster contained continuous time-series data from 2007 to 2020. Similarly, the CO2 emissions of South Korea were obtained as a predictable unidirectional causal variable for GTemp from GC analysis. Based on data from the past 34 years, significant causality (p-value = 0.01) was observed with the integrated effect of Pop, GDP, and CO2 on GTemp.ConclusionThis study patterned the time-series data using SOM and examined the causal relationship between variables using the GC test. The study framework could be used as a reference by future scholars, ecologists, and the United Nations Sustainable Development Goals.

Short-term Photovoltaic Power Forecasting Using SOM-based Regional Modelling Methods

The inherent intermittency and uncertainty of photovoltaic (PV) power generation impede the development of grid-connected PV systems. Accurately forecasting PV output power is an effective way to address this problem. A hybrid forecasting model that combines the clustering of a trained self-organizing map (SOM) network and an optimized kernel extreme learning machine (KELM) method to improve the accuracy of short-term PV power generation forecasting are proposed. First, pure SOM is employed to complete the initial partitions of the training dataset; then the fuzzy c-means (FCM) algorithm is used to cluster the trained SOM network and the Davies-Bouldin index (DBI) is utilized to determine the optimal size of clusters, simultaneously. Finally, in each data partition, the clusters are combined with the KELM method optimized by differential evolution algorithm to establish a regional KELM model or combined with multiple linear regression (MR) using least squares to complete coefficient evaluation to establish a regional MR model. The proposed models are applied to one-hour-ahead PV power forecasting instances in three different solar power plants provided by GEFCom2014. Compared with other single global models, the root mean square errors (RMSEs) of the proposed regional KELM model are reduced by 52.06% in plant 1, 54.56% in plant 2, and 51.43% in plant 3 on average. Such results demonstrate that the forecasting accuracy has been significantly improved using the proposed models. In addition, the comparisons between the proposed and existing state-of-the-art forecasting methods presented have demonstrated the superiority of the proposed methods. The forecasts of different methods in different seasons revealed the strong robustness of the proposed method. In four seasons, the MAEs and RMSEs of the proposed SF-KELM are generally the smallest. Moreover, the <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$R^{2}$</tex> value exceeds 0.9, which is the closest to 1.

Impulsive Noise Suppression Methods Based on Time Adaptive Self-Organizing Map

Removal of noise and restoration of images has been one of the most interesting topics in the field of image processing in the past few years. Existing filter-based methods can remove image noise; however, they cannot preserve image quality and information such as lines and edges. In this article, various classifiers and spatial filters are combined to achieve desirable image restoration. Meanwhile, the time adaptive self-organizing map (TASOM) classifier is more emphasized in our feature extraction and dimensionality reduction approaches to preserve the details during the process, and restore the images from noise. The TASOM was compared with the self-organizing map (SOM) network, and a suitable noise reduction method for images was attempted. As a result, we achieved an optimum method to reduce impulsive noise. In addition, by using this neural network, better noise suppression was achieved. Experimental results show that the proposed method effectively removes impulse noise and maintains color information as well as image details.

Iris Recognition Using Hybrid Self-Organizing Map Classifier and Daugman’s Algorithm

One of the neural network algorithms that can be used in iris recognition is self-organizing map (SOM). This algorithm has a weakness in determining the initial weight of the network, which is generally carried out randomly, which can result in a decrease in accuracy when an incorrect determination is made. The solution that is often used is to apply a hybrid process in determining the initial weight of the SOM network. This study takes an approach using the cosine similarity equation to determine the initial weight of the network SOM in order to increase recognition accuracy. In addition, the localization process needs to be carried out to limit the area of the iris image being studied so that it is easy for the recognition process to be carried out. The method proposed in this study for iris recognition, namely hybrid SOM and Daugman’s algorithm, has been tested on several people by capturing the iris of the eye using a digital camera. The captured eyes have been localized first using the Daugman’s algorithm, and then the image features were extracted using the GLCM and LBP methods. In the final stage of the study, an iris recognition comparison test was performed, and the results obtained an accuracy of 85.50% using the proposed method and an accuracy of 73.50% without performing a hybrid process on the SOM network.&#x0D;

Generating accurate negative samples for landslide susceptibility mapping: A combined self-organizing-map and one-class SVM method

The accuracy of data-driven landslide susceptibility mapping (LSM) is closely affected by the quality of non-landslide samples. This research proposes a method combining a self-organizing-map (SOM) and a one-class SVM (SOM-OCSVM) to generate more reasonable non-landslide samples. We designed two steps: first, a random selection, a SOM network, a one class SVM model, and a SOM-OCSVM model were used to generate non-landslide sample datasets. Second, four machine learning models (MLs)—namely logistic regression (LRG), multilayer perceptron (MLP), support vector machine (SVM), and random forest (RF)—were used to verify the effects of four non-landslide sample datasets on LSM. From the region along the Sichuan-Tibet Highway, we selected 11 conditioning factors and 1186 investigated landslides to perform the aforementioned experiments. The results show that the SOM-OCSVM method achieves the highest AUC (&amp;gt;0.94) and minimum standard deviation (&amp;lt;0.081) compared with other methods. Moreover, RF achieves the best performance in different datasets compared with other ML models. The aforementioned results prove that the proposed method can enhance the performance of ML models to produce more reliable LSM.

An Improved Ball K-Means Clustering Method Based on SOM

With the increase of the dimension and quantity of sample data, the calculation cost of K-Means clustering algorithm increases sharply. Therefore, a novel accelerated accurate K-Means clustering algorithm, called "Ball K-Means", has recently been used to reduce the computational cost. Although Ball K-Means reduces the computational cost, both this algorithm and K-Means algorithm lack the global search capability. K-means algorithm may fall into local minima because of its dependence on the initial center. The proper selection of the initial center vector becomes the key to improve the K-means algorithm. Therefore, self-organizing map (SOM) can be used to cluster and determine the clustering range quickly, and then the result can be used as the initial center vector of K-means method. Aiming at the problems that the initial clustering center of Ball K-Means algorithm is randomly selected in the stage of clustering calculation, and the clustering result may fall into a local optimal solution, this article uses SOM network to preliminarily process the data to obtain the initial clustering center of Ball K-Means algorithm, which significantly improves the clustering effect of the algorithm. Taking intrusion detection as an example, the effectiveness and superiority of the algorithm are verified by experiments.

A feature-level degradation measurement method for composite health index construction and trend prediction modeling

Accurate measurement of degradation levels and evolution trend for mechanical equipment has great significance to achieve the goal of condition-based maintenance (CBM) and increase the economic benefit. In this paper, a feature-level degradation measurement method is proposed for composite health index (CHI) construction and trend prediction, systematically blending the multi-source signals analysis theory and deep learning technology. First, a CHI is innovatively constructed based on a novel hybrid stacked auto-encoder (HSAE) and self-organizing map (SOM) network to comprehensively evaluate the degradation levels, which achieves the feature-level data fusion for degradation measurement. Second, considering the significant characteristics of strong nonlinear and random fluctuations for the generated CHI sequence, a hybrid algorithm, combining the variational mode decomposition (VMD) and a dynamic step size-based fruit fly optimization algorithm (DSSFOA), is applied for effectively extracting the multiscale frequency components from raw sequence and capturing the inherent laws of health degradation from different scales. Specifically, by designing a hierarchical search structure with dynamic step sizes in DSSFOA, the parameters of VMD can be optimized with high efficiency and accuracy. Finally, a scale-adaptive attention (SAA)-based bidirectional long short-term memory (BiLSTM) network and the corresponding dual-stage training strategy are developed to adaptively estimate the attention weights of different components and further predict the degradation trend in future. Extensive experiments on the C-MAPSS turbofan engine dataset and the PSU dataset validate the more superior performance of the proposed method on both of health index construction and degradation trend prediction than other existing methods.

Analysis of long term water quality variations driven by multiple factors in a typical basin of Beijing-Tianjin-Hebei region combined with neural networks

Water environment can be affected by multiple natural and human factors.This study aimed to explore the impacts on water quality from various driving factors including land use, landscape, social economy and climate in a typical basin of Beijing-Tianjin-Hebei Region combined with neural networks like self-organizing map and back propagation artificial neural network based on the water quality monitoring data.And redundancy analysis as well as gray correlation analysis were also adopted to explore the relationships between water quality and landscape indices as well as socioeconomic factors, respectively. The results showed that different land uses had different impacts on water quality. Agricultural additions in cropland had a great impact on surrounding water environment. The forest and grassland purified water to some degree, while the vegetation purification became weaker when there was serious pollution. The fragmentated landscape patches with high patch density caused by human activities also worsened water quality because its retention of pollutants was weakened. Besides, the per capita disposable income of rural residents and the per capita disposable urban income had the highest correlation degrees with most water quality indexes in all socioeconomic factors. In addition, temperature rise would promote dissolved oxygen consumption and eutrophication. Increasing precipitation brought pollutants into water via rainfall contains, nutrient transportation and soil erosion. The main conclusions were that both agricultural activities and urban development had impacts on water environment, and the vegetation purification became less obvious in heavily polluted areas. In addition, the rising temperature and precipitation disturbed water quality under the background of global warming. This study is helpful for water pollution management and it will give enlightenment for the water environment protection under the combining effects of human disturbance, economic development and climate change.

Static and Dynamic Joint Analysis for Operation Condition Division of Industrial Process With Incremental Learning

With the development of information and communication technologies, industrial cyber–physical systems (ICPSs) have accumulated a large amount of data, which enables us to convert data into industrial insight. However, since the industrial process of ICPS is always complicated and large scale, the raw data only contain a few operation condition information, which brings challenges to process monitoring and control. Thus, an efficient operation condition division method for ICPS is necessary. Although many operation condition division methods have been proposed, they were mainly relying on the static characteristics and ignored how the industrial process varies dynamically. Meanwhile, with the industrial process running, there may exist some new operation conditions that make the operation condition division task even more difficult. In order to grasp the static and dynamic features simultaneously of the industrial process and divide operation conditions accurately, we proposed an operation condition division method based on joint static and dynamic analysis with incremental learning. In detail, the slow feature analysis (SFA) and self-organizing map (SOM) network were proposed to extract the static and dynamic features jointly. Then, a division strategy was proposed to distinguish the operation condition changing points. For the new operating condition, we designed an incremental learning method based on the SOM network, which can update the operation condition model in real time. Extensive experiments, including a numerical simulation, two benchmark processes, and an industrial roasting process demonstrate that the proposed method can identify the operation conditions of the raw data in ICPS accurately and efficiently.

基于自组织映射网络解决旅行商问题的算法

基于自组织映射网络解决旅行商问题的算法

Tracking the Connection Between Brazilian Agricultural Diversity and Native Vegetation Change by a Machine Learning Approach

In Brazil, agribusiness has a considerable role in the country's GDP. Because of this, the State needs territorial planning to minimize the impacts on natural resources, especially in the Pantanal and Amazon biomes, where agribusiness has expanded. The lower the agricultural diversification, the lower the pattern of land use homogeneity, generally associated with agribusiness, especially when it occupies large areas with more technological productive units. This paper investigates the relationship between spatial diversification patterns and the dynamics of native vegetation in Brazil. We propose a feature engineering and clustering approach for 5570 Brazilian municipalities between 1999 and 2018. It was based on the unsupervised artificial neural network Self-Organizing Map (SOM) to divide the municipalities into homogeneous groups of agricultural products diversity trends. The results were compared with the change in vegetation area using data from the national land use-mapping project called Mapbiomas. The analysis allowed the identification of three different regimes of modification in native vegetation, particularly related to municipalities in Brazil's Midwest and North regions, indicating substantial changes in the Cerrado and Amazon biomes.

A Study on Different Deep Learning Algorithms Used in Deep Neural Nets: MLP SOM and DBN.

Deep learning is a wildly popular topic in machine learning and is structured as a series of nonlinear layers that learns various levels of data representations. Deep learning employs numerous layers to represent data abstractions to implement various computer models. Deep learning approaches like generative, discriminative models and model transfer have transformed information processing. This article proposes a comprehensive review of various deep learning algorithms Multi layer perception, Self-organizing map and deep belief networks algorithms. It first briefly introduces historical and recent state-of-the-art reviews with suitable architectures and implementation steps. Moreover, the various applications of those algorithms in various fields such as wireless networks, Adhoc networks, Mobile ad-hoc and vehicular ad-hoc networks, speech recognition engineering, medical applications, natural language processing, material science and remote sensing applications, etc. are classified.

A method for semantic-based image retrieval using hierarchical clustering tree and graph

Semantic extraction for images is an urgent problem and is applied in many different semantic retrieval systems. In this paper, a semantic-based image retrieval (SBIR) system is proposed based on the combination of growth partitioning tree (GP-Tree), which was built in the authors’ previous work, with a self-organizing map (SOM) network and neighbor graph (called SgGP-Tree) to improve accuracy. For each query image, a similar set of images is retrieved on the SgGP-Tree, and a set of visual words is extracted relying on the classes obtained from mask region-based convolutional neural networks (R-CNN), as the basis for querying semantic of input images on ontology by simple protocol and resource description framework query language (SPARQL) query. The experiment was performed on image datasets, such as ImageCLEF and MS-COCO, with precision values of 0.898453 and 0.875467, respectively. These results are compared with related works on the same image dataset, showing the effectiveness of the methods proposed.

Key Information Extraction Algorithm of Different Types of Digital Archives for Cultural Operation and Management.

In order to improve the effect of key information extraction from digital archives, a key information extraction algorithm for different types of digital archives is designed. Preprocess digital archive information, taking part of speech and marks as key information. Self-organizing feature mapping network is used to extract the key information features of digital archives, and the semantic similarity calculation results are obtained by combining the feature extraction results. Combine with mutual information collection, take that word with the highest mutual information value as the collection cent, traverse all keywords, and take the central word as the key information of digital archives to complete the extraction of key information. Experiments show that the recall rate of the algorithm ranges from 96% to 99%, the extraction accuracy of key information of digital archives is between 96 and 98%, and the average extraction time of key information of digital archives is 0.63 s. The practical application effect is good.

Long Short-Term Memory Neural Network with Transfer Learning and Ensemble Learning for Remaining Useful Life Prediction.

Prediction of remaining useful life (RUL) is greatly significant for improving the safety and reliability of manufacturing equipment. However, in real industry, it is difficult for RUL prediction models trained on a small sample of faults to obtain satisfactory accuracy. To overcome this drawback, this paper presents a long short-term memory (LSTM) neural network with transfer learning and ensemble learning and combines it with an unsupervised health indicator (HI) construction method for remaining-useful-life prediction. This study consists of the following parts: (1) utilizing the characteristics of deep belief networks and self-organizing map networks to translate raw sensor data to a synthetic HI that can effectively reflect system health; and (2) introducing transfer learning and ensemble learning to provide the required degradation mechanism for the RUL prediction model based on LSTM to improve the performance of the model. The performance of the proposed method is verified by two bearing datasets collected from experimental data, and the results show that the proposed method obtains better performance than comparable methods.

Adaptive overheating cover for a solar water heater

Abstract Solar water heating systems have been widely used around the world. However, exposure to sunlight can overheat the device, affecting the efficiency and durability of the system. This article proposes an adaptive deck controller that protects the system from overheating without compromising the availability of domestic hot water. Solar water heaters are considered one of the most effective ways to reduce a home’s carbon footprint. They are a renewable energy source that reduces reliance on fossil fuels and saves money. Thus, this paper aims to develop a dynamic cover for solar water heaters that prevent overheating using an artificial neural network to optimize the design of control systems. Based on a self-organizing map network, the controller automatically adjusts the temperature of the solar collector through a fabric screen covering the main subsystems, depending on many parameters such as weather conditions, collector temperature and domestic hot water depending on demand. A suggested technique of four different shade percentages (0%, 20%, 25% or 32%) can avoid overheating and maintain the amount of hot water the home needs. Although renewable energy is free, proper controls are required to ensure maximum efficiency or proper use. In addition, the control of renewable energy leads to longer service life.