Kohonen Self-organizing Neural Network Research Articles

Assessment of opportunities for the development of organic agriculture in municipal areas

Purpose: of the research is to develop a methodology for assessing the development opportunities of organic agriculture in municipal areas based on their clustering.Methods: within the framework of the study, methods of comparing averages, grouping, clustering based on Kohonen's self-organizing neural network were used. The study was conducted on the materials of the Penza region, the complex of analyzed indicators was formed on the basis of official statistical data on municipal districts of the region. The applied approach assumes the possibility of carrying out similar calculations based on the materials from other subjects of the Russian Federation.Results: the study proposes a system of indicators characterizing the development potential of the production of certain types of organic agricultural products on the territory of municipal districts. Clustering of districts was carried out, as a result, 6 clusters were identified, characterized by various combinations of the values of the analyzed factors. The characteristics of each cluster in terms of opportunities for the development of organic agriculture are given. The areas of activity with the best potential are indicated, as well as the main unfavorable factors. 2 clusters have been identified that can be considered as pilot ones for the implementation of the regional program for the development of organic agriculture.Conclusions and Relevance: at the initial stage of the development of the organic production sector, it is important to identify growth points, types of agricultural activities in certain areas that may be the most promising within the framework of the introduction of the principles of organic farming. Given the diversity and high variability of the values of indicators that determine the potential for the development of organic agriculture in individual territories, clustering can be an effective method that allows differentiating and increasing the targeting of state support measures for the relevant producers. The application of efforts for the development of organic agriculture, taking into account the peculiarities of the formed clusters, will be more effective.

The Use of Neural Network Modeling Methods to Determine Regional Threshold Values of Hydrochemical Indicators in the Environmental Monitoring System of Waterbodies.

The regulation of the anthropogenic load on waterbodies is carried out based on water quality standards that are determined using the threshold values of hydrochemical indicators. These applied standards should be defined both geographically and differentially, taking into account the regional specifics of the formation of surface water compositions. However, there is currently no unified approach to defining these regional standards. It is, therefore. appropriate to develop regional water quality standards utilizing modern technologies for the mathematical purpose of methods analysis using both experimental data sources and information system technologies. As suggested by the use of sets of chemical analysis and neural network cluster analysis, both methods of analysis and an expert assessment could identify surface water types as well as define the official regional threshold values of hydrochemical system indicators, to improve the adequacy of assessments and ensure the mathematical justification of developed standards. The process for testing the proposed approach was carried out, using the surface water resource objects in the territory of the Republic of Tatarstan as our example, in addition to using the results of long-term systematic measurements of informative hydrochemical indicators. In the first stage, typing was performed on surface waters using the neural network clustering method. Clustering was performed based on sets of determined hydrochemical parameters in Kohonen's self-organizing neural network. To assess the uniformity of data, groups in each of the selected clusters were represented by specialists in this subject area's region. To determine the regional threshold values of hydrochemical indicators, statistical data for the corresponding clusters were calculated, and the ranges of these values were used. The results of testing this proposed approach allowed us to recommend it for identifying surface water types, as well as to define the threshold values of hydrochemical indicators in the territory of any region with different surface water compositions.

Autism Spectrum Disorder Diagnosis Using Ensemble ML and Max Voting Techniques

Difficulty in communicating and interacting with other people are mainly due to the neurological disorder called autism spectrum disorder (ASD) diseases. These diseases can affect the nerves at any stage of the human being in childhood, adolescence, and adulthood. ASD is known as a behavioral disease due to the appearances of symptoms over the first two years that continue until adulthood. Most of the studies prove that the early detection of ASD helps improve the behavioral characteristics of patients with ASD. The detection of ASD is a very challenging task among various researchers. Machine learning (ML) algorithms still act very intelligent by learning the complex data and predicting quality results. In this paper, ensemble ML techniques for the early detection of ASD are proposed. In this detection, the dataset is first processed using three ML algorithms such as sequential minimal optimization with support vector machine, Kohonen self-organizing neural network, and random forest algorithm. The prediction results of these ML algorithms (ensemble) further use the bagging concept called max voting to predict the final result. The accuracy, sensitivity, and specificity of the proposed system are calculated using confusion matrix. The proposed ensemble technique performs better than state-of-the art ML algorithms.

Efficiency of the t-distribution stochastic neighbor embedding technique for detailed visualization and modeling interactions between agricultural soil quality indicators

Dimensionality reduction is important for revealing important details that may be useful in decision-making. Although different dimensionality reduction methods have been applied in several soil-based studies, Kohonen self-organizing map neural network (KSOM-NN) has attracted significant attention from researchers because of the quality of data visualization and interpretation. However, there is a dearth of studies that compare KSOM-NN and other robust data reduction techniques such as the t -distribution stochastic neighbor embedding ( t -SNE) method to improve visualization and interpretation of the relationships between soil quality indicators in agricultural soil. This study compares the above-mentioned methods for characterizing soil quality indicators including particle size distribution, soil organic matter (SOM), cation exchange capacity (CEC), soil reaction (pH), electrical conductivity (EC), zinc (Zn), iron (Fe), manganese (Mn), potassium (K) and phosphorus (P) in agricultural dryland. There were strongly positive associations identified between some of the variables studied for example, clay/Fe (r = 0.95), clay/SOM (r = 0.79) and Mn/Zn (r = 0.90) based on the correlation matrix output. According to the KSOM-NN, the best map size was a 4 by 7 with Quantization error (QE) = 0.108, Topographic error (TE) = 0.875 and Kaski-Lagus error (K-LE) = 9.104. This map only yielded 2 main clusters. As for the t -SNE, applying various perplexity values (i.e. 5, 6, 7, 8, 9 and 10) enabled better visualization of the soil quality indicators as observed from the cluster formations than the KSOM-NN. Ultimately, the t -SNE was considered a better and promising method for assessing the interactions of soil quality indicators and has the potential to appropriately visualise as well as improve the interpretability of soil results by identifying essential features and similarities. The results of this study have good implications for agricultural soil management and decision-making. Moreover, the results are preliminary findings that may be used in the future to verify detailed aspects of soil biogeochemistry within the study area. • t -SNE outperformed KSOM-NN in soil properties dimensionality reduction and visualization. • The t -SNE method is rarely used for data reduction and visualization in soil-based research. • More detailed clusters of the soil quality indicators were observed while using the t -SNE compared to KSOM-NN. • The t -SNE is recommended for future applications in soil quality characterization research.

Research on Model and Algorithm of User Access Pattern Data Mining

The era of big data has come. Today, all kinds of information and data are showing explosive growth. Internet activities of different scales are struggling to catch up with the pace and pace of the development of "big data. This paper uses time series mining algorithms, and uses Microsoft's data mining tools to model the data sets collected from data halls, so as to discover the user's online behaviour patterns and potential online rules within a certain period of time. we made reasonable suggestions for the scientific management of the campus network. A new clustering method based on the improved Kohonen self-organizing feature mapping neural network is proposed. A Gaussian-shaped membership function has introduced to output several neurons with a degree of membership greater than the threshold, thereby solving the problem of mining users' multiple interests.

An AVMD Method Based on Energy Ratio and Deep Belief Network for Fault Identification of Automation Transmission Device

Considering that the vibration signals of gears and bearings in the automatic transmission device are complex and the fault features are difficult to extract. This paper proposes a method for extracting fault features of transmission device using adaptive variational modal decomposition (AVMD), and uses deep belief network (DBN) for pattern recognition. The vibration signal is decomposed by AVMD using energy ratio method. The intrinsic mode function (IMF) with abundant fault information is obtained. By calculating the energy entropy of each IMF component and form a high-dimensional feature vector as the input of DBN to establish an early fault identification model. The early fault data of the PHM2009 transmission device experimental platform was selected for identification and analysis. The identification results show that AVMD can extract the weak features of transmission device fault signals more accurately than empirical mode decomposition (EMD). Moreover, DBN has a higher fault identification accuracy rate than support vector machine (SVM), probabilistic neural network (PNN), back propagation neural network (BP) and Kohonen self-organizing competition neural network.

Integrating machine learning with QFD for selecting functional requirements in construction automation

An important aspect of the conceptual design is at the customer requirement definition stage, where an optimal number of functional requirements are specified with the application of quality function deployment. To facilitate a systematic specification of functional requirements, state-of-the-art unsupervised machine learning techniques will be introduced in the feature selection of functional requirements. However, the scarcity of references on unsupervised feature selection in the literature reflects the difficulty associated with this topic. At the customer requirement definition phase, three techniques will be proposed for selecting functional requirements, namely: (a) principal component analysis, (b) forward orthogonal search, and (c) Kohonen self-organizing map neural network. These machine learning feature selection techniques address the limitations of current approaches in systematically determining the minimum functional requirements from the mapping of customer requirements in quality function deployment. When applied to the conceptual design of the transportable automated wood wall framing machine that is under development at the University of Alberta, the proposed feature selection techniques have been observed to be: (i) fast, (ii) amenable to small quality function deployment dataset, and (iii) adequate in realizing design objectives. The results presented in this paper can be easily extended to online determination of customer requirements and functional requirements, project management, contract management, and marketing.

Evaluation of models to predict the influence of chemical pretreatment on the peels of Nephelium lappaceum L. based on pyrolysis kinetic parameters obtained using a combined Fraser-Suzuki function and Friedman’s isoconversional method

In the present work, two chemometric strategies were evaluated to assess the influence of the chemical pretreatment: a Kohonen self-organizing map neural network and a Support Vector Machine Learning algorithm. Kohonen ANN was able to predict the influence of chemical treatment mainly on pseudo-component lignin, since cellulose and hemicellulose have similar structures; presenting a topological error of 0.044 and a quantification error of 0.193, while the Vector Supporting Machine (SVM) algorithm was able to predict the classes with accuracy of 0.956, area under the ROC curve (AUC) of 0.992 and F1 score of 0.956, proving to be a more effective strategy in predicting the influence of the chemical treatment on the rambutan shells based on the kinetic parameters. It was also observed an increase in the activation energy for the cellulose pretreated with sodium hydroxide and with phosphoric acid, mainly due to the partial conversion of type I cellulose into type II cellulose, ranging from 152 to 155 kJ/mol versus 124 kJ/mol for the untreated sample. The thermal decomposition mechanism did not change significantly, being estimated as pseudo-zero order, using the Master-Plot method. TG-FTIR analysis reveals that the main gases emitted during pyrolysis are CO2, H2O and CO, being the carbon dioxide-related band an indicator of the influence of the chemical treatment. This analysis also reveals the release of sulfur compounds from the sample treated with sulfuric acid, indicating that the environmental point of view of this sample shows no viability as use as pyrolysis feedstock.

Phenotyping of masked hypertension based on the clustering of 24-hour blood pressure monitoring data

Aim. Phenotyping of masked hypertension (MH) based on clustering of 24-hour ambulatory blood pressure monitoring (ABPM) results data and comparing it with indicators of target organ damage.Material and methods. We retrospectively analyzed medical records of 207 men with a median age of 34,6 years and normal office blood pressure (BP), who were systematically exposed to occupational stressors and were undergoing a routine in-patient examination. All patients underwent ABPM, echocardiography, carotid ultrasound, glomerular filtration rate estimation. The clustering of ABPM data was carried out using the Kohonen self-organizing neural networks and K-means algorithm. Data processing was performed in the R programming language using the RStudio environment.Results. MH was diagnosed in 142 (68,6%) patients which were divided into 3 clusters according to 4 criteria factors of ABPM: systolic-diastolic (SDMH) — 50,7%, isolated systolic (ISMH) — 27,5% and isolated diastolic (IDMH) — 21,8%. Majority (51,4%) of patients with SDMH were characterized by a relatively balanced distribution of episodes of increased systolic and diastolic BP during the day, and other ratios were much less common. In patients with ISMH, hypertension prevailed in the daylight, and in IDMР — nighttime. Individual clusters differed significantly in terms of the 24-hour BP profile and signs of target organ damage. Various types of cardiac remodeling were recorded in 24 (17%) patients with MH, 16 (66,7%) of which belonged to the SDMH phenotype. Intima-media thickening >0,9 mm occurred in 40% of patients with IDMH, in 30% — SDMH and only 9% — ISMH. At the same time, the majority of patients with glomerular hyperfiltration was recorded in patients with ISMH (20,5%), and with hypofiltration — IDMH (29%) and SDMH (23,6%).Conclusion. Modern technologies of clustering increase the effectiveness of risk stratification for patients with MH, and contribute to the personification of preventive and therapeutic programs.

Analysis of the stability factors of Ukrainian banks during the 2014–2017 systemic crisis using the Kohonen self-organizing neural networks

The article proposes an approach to analyzing reliability factors of commercial banks during the 2014–2017 systemic crisis in the Ukrainian banking system, using the Kohonen self-organizing neural networks and maps. As a result of an experimental study, data were obtained on financial factors affecting the stability of a commercial bank in a crisis period. It has been concluded that during the banking crisis in Ukraine in 2014–2017, the resource base of a bank was the main factor of this bank stability. The most preferred sources of resources were funds from other banks (bankruptcy rate of 5.7%) and legal entities (bankruptcy rate of 8%), and the least stable were funds from individuals (bankruptcy rate of 28.5%). The relationship between financial stability and the amount of capital and the structure of bank loans is less pronounced. However, one can say that banks that focused on lending to individuals experienced a worse crisis than banks whose main borrowers were legal entities. The tools considered in the article (the Kohonen self-organizing neural networks and maps) allow for efficiently segmenting data samples according to various criteria, including bank solvency. The “hazardous” zones with a high bankruptcy rate (up to 49.2%) and the “safe” zone with a low rate of bankruptcy (6.3%) were highlighted on the map constructed. These results are of practical value and can be used in analyzing and selecting counterparties in the banking system during a downturn.

Prediction of Transition Temperature of Alloy steel Austenite Based on Self-Organizing Neural Network

In order to determine the key chemical elements of the austenite transformation temperature (Ac1/Ac3) of alloy steel, this paper uses stepwise regression so that the number of neurons in the input layer of Kohonen self-organizing neural network is also determined and the network structure will be optimized. The learning algorithm is improved by combining non-tutor learning and supervised learning. Based on the optimized Kohonen neural network, an Ac1/Ac3 transition temperature predictive model was established. The prediction accuracy of this model which is based on Kohonen self-organizing neural network is higher in tests.The relative error of Ac1 is less than 3.01%, and the relative error of Ac3 is also below3.02%, which is significantly better than the prediction accuracy of stepwise regression. With predicting the critical temperature of Ac1/Ac3 based on Kohonen Network, the actual heating temperature can be determined, which has important practical application value in both steel heat treatment to guarantee quality and shortening the physical experiment period.CLC number: TP391 Document code: A

Application of kohonen neural networks to search for regions of interest in the detection and recognition of objects

One of the most effective ways to improve accuracy and speed of recognition algorithms is to preliminary distinguish the regions of interest in the analyzed images. We studied a possibility of application of self-organizing maps and a Kohonen neural network for detection of regions of interest at a radar or satellite image of underlying surface. There is a high probability of finding an object of interest for further analysis in the found regions of interest. The definition of region of interest is necessary most of all to automate and speed up the process of search and recognition of objects of interest. The relevance is due to the increasing number of satellites. The study presents the process of modeling, analysis and comparison of the results of application of these methods for determination of regions of interest in recognition of images of aircraft against the background of underlying surface. It also describes the process of preliminary processing of input data. The study presents a general approach to construction and training of the Kohonen self-organizing map and neural network. Application of Kohonen maps and neural network makes it possible to decrease an amount of data analyzed by 15–100 times. It speeds up the process of detection and recognition of an object of interest. Application of the above algorithm reduces significantly the required number of training images for a convolutional network, which performs the final recognition. The reduction of a training sample occurs because the size of parts of an input image supplied to the convolutional network is bounded with the scale of an image and it is equal to the size of the largest detected object. Kohonen neural network showed itself more efficient in relation to this task, since it places cluster centers on the underlying surface rarely due to independence of weight of neurons on neighboring centers. These technical solutions could be used in the analysis of visual data from satellites, aircraft, and unmanned cars, in medicine, robotics, etc.

Pattern recognition of seismogenic nodes using Kohonen self-organizing map: example in west and south west of Alborz region in Iran

Pattern recognition of seismic and morphostructural nodes plays an important role in seismic hazard assessment. This is a known fact in seismology that tectonic nodes are prone areas to large earthquake and have this potential. They are identified by morphostructural analysis. In this study, the Alborz region has considered as studied case and locations of future events are forecast based on Kohonen Self-Organized Neural Network. It has been shown how it can predict the location of earthquake, and identifies seismogenic nodes which are prone to earthquake of M5.5+ at the West of Alborz in Iran by using International Institute Earthquake Engineering and Seismology earthquake catalogs data. First, the main faults and tectonic lineaments have been identified based on MZ (land zoning method) method. After that, by using pattern recognition, we generalized past recorded events to future in order to show the region of probable future earthquakes. In other word, hazardous nodes have determined among all nodes by new catalog generated Self-organizing feature maps (SOFM). Our input data are extracted from catalog, consists longitude and latitude of past event between 1980–2015 with magnitude larger or equal to 4.5. It has concluded node D1 is candidate for big earthquakes in comparison with other nodes and other nodes are in lower levels of this potential.

Análise exploratória de simulações sociais computacionais por meio de estatística multivariada e mapas auto-organizáveis

Este trabalho teve como objetivo a aplicação da análise de componentes principais, da análise de agrupamentos pelo método hierárquico aglomerativo e da rede neural do tipo Mapa Auto-Organizável de Kohonen na análise exploratória de resultados de simulações sociais computacionais do sistema socioterritorial “Território Rural Sul Sergipano”. A metodologia basea-se na Sociologia da Ação Organizada e no método Soclab. As análises estatísticas mostraram que as simulações do sistema socioterritorial geraram resultados que confirmam a tendência do sistema à estabilidade. No entanto, observou-se, por meio da rede neural, que há diferenças, ainda que pequenas, entre as situações que levam o sistema à estabilidade.

An efficient initialization mechanism of neurons for Winner Takes All Neural Network implemented in the CMOS technology

The paper presents a new initialization mechanism based on a Convex Combination Method (CCM) for Kohonen self-organizing Neural Networks (NNs) realized in the CMOS technology. A proper selection of initial values of the neuron weights exhibits a strong impact on the quality of the overall learning process. Unfortunately, in case of real input data, e.g. biomedical data, proper initialization is not easy to perform, as an exact data distribution is usually unknown. Bad initialization causes that even 70%–80% of neurons remain inactive, which increases the quantization error and thus limits the classification abilities of the NN. The proposed initialization algorithm has a couple of important advantages. Firstly, it does not require a knowledge of data distribution in the input data space. Secondly, there is no necessity for an initial polarization of the neuron weights before starting the learning process. This feature is very convenient in case of transistor level realizations. In this case the programming lines, which in other approaches occupy a large chip area, are not required. We proposed a modification of the original CCM algorithm. A new parameter which in the proposed analog CMOS realization is represented by an external current, allows to fit the behavior of the mechanism to NNs containing different numbers of neurons. The investigations show that the modified CCM operates properly for the NN containing even 250 neurons. A single CCM block realized in the CMOS 180nm technology occupies an area of 300μm2 and dissipates an average power of 20μW and at data rate of up to 20MHz.

Biosom: gene synonym analysis by self-organizing map.

There are several guidelines for gene nomenclature, but they are not always applied to the names of newly identified genes. The lack of standardization in naming genes generates inconsistent databases with errors such as genes with the same function and different names, genes with different functions and the same name, and use of an abbreviated name. This paper presents a methodology for predicting synonyms in a given gene nomenclature, thereby detecting and minimizing naming redundancy and inconsistency and facilitating the annotation of new genes and data mining in public databases. To identify gene synonyms, i.e., gene ambiguity, the methodology proposed begins by grouping genes according to their names using a Kohonen self-organizing map artificial neural network. Afterwards, it identifies the groups generated employing the Matrix-U technique. The employment of such techniques allows one to infer the synonyms of genes, to predict probable hypothetical gene names and to point out possible errors in a database record. Many mistakes related to gene nomenclature were detected in this research, demonstrating the importance of predicting synonyms. The methodology developed is applicable for describing hypothetical, putative and other types of genes without a known function. Moreover, it can also indicate a possible function for genes after grouping them.

VYUŽITÍ UMĚLÝCH NEURONOVÝCH SÍTÍ PRO KLASIFIKACI SIGNÁLŮ BIOSENZORŮ

Biosensors are analytical devices that transforms chemical information, ranging from the concentration of a specific sample component to total composition analysis, into an analytical signal and that utilizes a biochemical mechanism for the chemical recognition. The complexity of biosensor construction and generation of measured signal requires the development of new method for signal eva­luation and its possible defects recognition. A new method based on artificial neural networks (ANN) was developed for recognition of characteristic behavior of signals joined with malfunction of sensor. New algorithm uses unsupervised Kohonen self-organizing neural networks. The work with ANN has two phases – adaptation and prediction. During the adaptation step the classification model is build. Measured data form groups after projection into two-dimensional space based on theirs similarity. After identification of these groups and establishing the connection with signal disorders ANN can be used for evaluation of newly measured signals. This algorithm was successfully applied for 540 signal classification obtained from immobilized acetylcholinesterase biosensor measurement of organophosphate and carbamate pesticides in vegetables, fruits, spices, potatoes and soil samples. From six different signal defects were successfully classified four – low response after substrate addition, equilibration at high values, slow equilibration after substrate addition respectively low sensitivity on syntostigmine.

Fuzzy Domain Image Segmentation Based on Space Information

Synthetic aperture radar (SAR) Image containing the coherent speckle noise, traditional methods can't very well to SAR image segmentation, based on SAR image grayscale distribution model, this paper proposes a combination of pixel gray level and area information of the image segmentation method. First of all, based on SAR image in all kinds of feature of gray histogram presents the facts of the Rayleigh distribution, fitting of image gray level of probability density distribution; Then use fuzzy biggest office methods select the optimal threshold, and the resulting statistics as the related parameters of image, the image blur ; Then to 3 x3 neighborhood pixels within the membership degree of fuzzy characteristics of Kohonen self-organizing map neural network is adopted to improve the coarse segmentation, the segmentation of noise and boundary classes pixels using a region growing technique relaxation iteration to fine segmentation.

AI-WSN: Adaptive and Intelligent Wireless Sensor Network

This paper proposes employment of artificial neural network techniques to develop in-network “intelligent computation” and “adaptation” capability for wireless sensor networks to improve their functionality, utility and survival aspects. The goal is to introduce computational intelligence capability for the wireless sensor networks to become adaptive to changes within a variety of operational contexts and to exhibit intelligent behaviour. The characteristics of wireless sensor networks bring many challenges, such as the ultra large number of sensor nodes, dense deployment, changing topology structure, and the most importantly, the limited resources including power, computation, storage, and communication capability. All these require the applications and protocols running on wireless sensor network to be not only energy-efficient, scalable and robust, but also “adapt” to changing environment or context, and application scope and focus among others, and demonstrate intelligent behaviour. Feasibility of the proposed approach is demonstrated through a simulation-based case study which entailed a clustering of Iris data using Kohonen's self-organizing map neural network which was embedded across a wireless sensor network.

Color reduction using a multi-stage Kohonen Self-Organizing Map with redundant features

Reducing the number of colors in an image while preserving its quality, is of importance in many applications such as image analysis and compression. It also decreases the memory and transmission bandwidth requirements. Moreover, classification of image colors is applicable in image segmentation, and object detection and separation, as well as producing pseudo-color images. In this paper, the Kohonen Self-Organizing Map Neural Network is employed to form an adaptive color reduction method. To enhance the performance of this method, we have used redundant features obtained by one-to-one functions from three main components of the color image (e.g. Red, Green and Blue channels). Exploiting these features will increase the color discrimination and details illustration ability of the network compared to the conventional approaches. This method leads to satisfactory results in image segmentation, especially in small object detection problems. It is also investigated that if the number of features in Kohonen network grows even by using non-deterministic one-to-one functions, the network revenue considerably improves. Moreover, we will study the effect of various adaptation algorithms in Kohonen network training stage. Again, using a multi-stage color reduction procedure which employs both Kohonen neural networks and conventional vector quantization schemes improves the performance. Several experimental results are represented to illustrate the characteristics of different approaches.