SOM Algorithms Research Articles

Cluster Head Selection in Device to Device (D2D) Communication Based on Weighted Network Performance Factor

In this study, an unsupervised machine learning algorithm, Self Organizing Map was utilized to cluster D2D User Equipment using their network values as inputs. A weighting factor referred to as Hardware Sensing Factor (HSF) was formulated to take into account the device’s channel quality and the status of its underlying hardware circuitry. The values of the HSF were used as inputs to cluster the devices and to select cluster head for each cluster. The performance of SOM when HSF was input was compared with the performance when RSSI, RSRP or RSRQ was used as input. The comparison showed that the use of HSF as input to SOM cluster algorithm gave better cluster performance than the use of respective network values such as RSSI, RSRP or RSRQ. In addition, the use of HSF as input data to both SOM and K-Means algorithms showed that SOM cluster formation has better performance than K-Means algorithm.

Vector batch SOM algorithms for multi-view dissimilarity data

Multi-view data has become fairly important since large amounts of information are constantly generated from different sources. So far, most multi-view research on unsupervised learning has focused on clustering algorithms suitable for analyzing vector data. Instead, our study presents two batch SOM algorithms for multi-view dissimilarity data, namely a batch SOM with each cluster representative as a vector of a subset of objects (vector of set-medoids) and another batch SOM for multi-view dissimilarity data with each cluster representative as a matrix whose elements are the weights of its objects regarding the dissimilarity matrices (weighted medoids). These algorithms are suitable for clustering and visualizing multi-view dissimilarity data. In addition, they provide the relevance weights on the importance of each dissimilarity matrix (view) concerning each cluster related to the neurons in the map. Furthermore, the weight can be computed both locally, for each cluster, or globally, for the whole partition. Experiments with 12 datasets showed the importance of multi-view algorithms with respect to single-view algorithms in some contexts. Among the multi-view algorithms, the results showed the need to consider the relevance weights of the dissimilarity matrices. Moreover, an application study on the FOREST TYPE dataset found that some of the proposed methods provide better-quality maps.

Methods of Pre-Clustering and Generating Time Series Images for Detecting Anomalies in Electric Power Usage Data

As electricity supply expands, it is essential for providers to predict and analyze consumer electricity patterns to plan effective electricity supply policies. In general, electricity consumption data take the form of time series data, and to analyze the data, it is first necessary to check if there is no data contamination. For this, the process of verifying that there are no abnormalities in the data is essential. Especially for power data, anomalies are often recorded over multiple time units rather than a single point. In addition, due to various external factors, each set of power consumption data does not have consistent data features, so the importance of pre-clustering is highlighted. In this paper, we propose a method using a CNN model using pre-clustering-based time series images to detect anomalies in time series power usage data. For pre-clustering, the performances were compared using k-means, k-shapes clustering, and SOM algorithms. After pre-clustering, a method using the ARIMA model, a statistical technique for anomaly detection, and a CNN-based model by converting time series data into images compared the methods used. As a result, the pre-clustered data produced higher accuracy anomaly detection results than the non-clustered data, and the CNN-based binary classification model using time series images had higher accuracy than the ARIMA model.

Non-Uniform Input-Based Adaptive Growing Neural Gas for Unstructured Environment Map Construction

The research and development of special robots such as excavation robots is an important way to achieve safe and efficient production in coal mines. Affected by the unstructured environment such as complex working conditions and unsteady factor disturbances, the real-time construction of section environment maps that can accurately describe the environment and facilitate trajectory planning and decision making has become a key scientific problem to be solved as soon as possible. Therefore, non-uniform input based adaptive growing neural gas for unstructured environment map construction has been proposed. Considering complex load identification, real-time location identification, and the types of unsteady disturbance factors and working conditions, a set of environment identification models has been established based on a large amount of underground measured data training. These models can express whether the section environment has changed, as well as the type and magnitude of the change, to realize the overall knowledge extraction and parametric representation of the unstructured environment. Then, in order to solve the problems of inaccurate topology, excessive aging of connecting edges, and excessive deletion of nodes in non-uniform input environment, an adaptive growing neural gas algorithm based on non-uniform input environment (AGNG-NU) is proposed. Featured by a dynamic response deletion mechanism and adaptive adjustment mechanism of neuron parameters, the generated nodes and their topology can be dynamically adjusted according to the density of regional sample points. Several sets of non-uniform input environments are set to test the algorithm. The experimental results show that the topological maps established by AGNG-NU express clearer environmental information and, at the same time, the accuracy and distribution are improved by 8% and 15%, respectively, compared with the basic GNG algorithm. The accuracy and the distribution have also been significantly improved compared with other common SOM and GCS algorithms.

Two weighted c-medoids batch SOM algorithms for dissimilarity data

This paper proposes two new batch SOM algorithms for dissimilarity data, namely RBSOM-CWMdd and RBSOM-ACWMdd, both designed to give a crisp partition aiming to preserve the topological properties of the data on the map. RBSOM-CWMdd is a batch SOM algorithm for dissimilarity data where each cluster representative is a set of weighted objects whose cardinality is fixed, being the same for all clusters. These weights are computed according to each object relevance to the referred cluster. Likewise, RBSOM-ACWMdd is a batch SOM algorithm for dissimilarity data where each cluster representative is a vector of weighted objects selected according to its relevance to the referred cluster. Therefore, the dimensionality of the cluster representatives self adapt to the particular dataset analysed, change at each algorithm iteration and can differ from one cluster to another. Experiments with 12 datasets from the UCI machine learning repository regarding the metrics of Normalized Mutual Information, Topological error, and Silhouette Coefficient showed that the proposed methods improved, respectively, the traditional set-medoids and multi-medoids SOM methods with a competitive temporal complexity. In addition, it was performed an application study on Ecoli dataset where the proposed RBSOM-ACWMdd algorithms produced a better mapping from a clustering point of view.

Color segmentation of multicolor porous printed fabrics by conjugating SOM and EDSC clustering algorithms

In order to achieve accurate color segmentation of printed fabrics, the color segmentation algorithm combining the self-organizing maps neural network and the efficient dense subspace clustering was proposed in this paper. After pre-processing of the fabric image, the primary clustering was implemented by the self-organizing maps algorithm, then the secondary clustering was done by the efficient dense subspace clustering algorithm. The optimal silhouette coefficient is introduced into the clustering process of the efficient dense subspace clustering algorithm to determine the number of clustering centers automatically. Finally, by the post-processing including gray-scale transformation, binarization and open operation, the mis-segmentation of edge color was eliminated, making the algorithm more suitable for industrial application. Experiments were carried out and results show that the algorithm proposed in this paper can recognize the color of small areas accurately and segment the color of complex printed fabric images. The color segmented results of 20 printed fabrics show that the accuracy of the algorithm proposed in this paper reaches 88.3%.

Development and validation of a system for the prediction of excipient-excipient incompatibility using machine learning tools

The importance of pharmaceutical excipients in the creation of any dosage form is critical. These excipients are occasionally to blame for product underperformance and dosage form deterioration. Product deterioration and underperformance could be attributed to incompatibilities between drug and excipient or sometimes excipient and excipient either due to the presence of reactive impurities in the excipients or a reaction between the functional groups present on the excipients. Although, the drug and excipient incompatibilities are monitored and reported, excipient-excipient incompatibilities are overlooked due to a paucity of literature. Pharmaceutical companies used to work in a controlled environment (compatibility tests between excipients to determine the best excipients for dosage form creation) and utilize mitigation measures to suppress any incompatibilities between excipients when necessary. In the current paper, we propose a system to predict excipient-excipient incompatibilities that can occur during dosage form development, possible incompatibility reactions, as well as a potential way to counter these incompatibilities by suggesting alternative excipients based on the descriptor knowledge of the excipients. The system was developed and validated using SOM and random forest algorithms. Two systems were developed based on incompatibility and property relationship, this approach provides a wide scope of our system to select compatible excipients in place of incompatible ones. Two systems predicted top 10 key incompatibilities and properties which cover the maximum range of the pharmaceutical research and development. After the validation using random forest, our system was able to find the structural components responsible for these top incompatibilities and properties. The prediction accuracy was found to be >85%. This was confirmed by preparing confusion matrix of the top incompatibilities and properties and assigning values of 0.5 to predict the outcome. The excipients with probability of ≥0.5 were considered active (possessing incompatibility or property) and the excipients with probability

Application of Supervised SOM Algorithms in Predicting the Hepatotoxic Potential of Drugs.

The hepatotoxic potential of drugs is one of the main reasons why a number of drugs never reach the market or have to be withdrawn from the market. Therefore, the evaluation of the hepatotoxic potential of drugs is an important part of the drug development process. The aim of this work was to evaluate the relative abilities of different supervised self-organizing algorithms in classifying the hepatotoxic potential of drugs. Two modifications of standard counter-propagation training algorithms were proposed to achieve good separation of clusters on the self-organizing map. A series of optimizations were performed using genetic algorithm to select models developed with counter-propagation neural networks, X-Y fused networks, and the two newly proposed algorithms. The cluster separations achieved by the different algorithms were evaluated using a simple measure presented in this paper. Both proposed algorithms showed a better formation of clusters compared to the standard counter-propagation algorithm. The X-Y fused neural network confirmed its high ability to form well-separated clusters. Nevertheless, one of the proposed algorithms came close to its clustering results, which also resulted in a similar number of selected models.

A Comparitive Analaysis of Fuzzy Particle Swarm Optimization with SOM and EM Algorithms

Clustering is a kind of unsupervised learning, the process of dividing a given data set into groups according to the similarity of a given data set, and similarity is performed according to distance. Some researchers have developed some data clustering algorithms, others have implemented new algorithms, and some have studied existing data and compared other data clustering algorithms. Here are some previous studies that considered the impact of several factors on the performance of specific data clustering algorithms and compared the results. However, this study is different from algorithms and factor analysis; this article aims to study and compare functional weighted fuzzy particle clustering optimization with self-configuration mapping and expectation maximized clustering algorithms. All of these algorithms, depending on the size of the data, the number of clusters, the type of data set, and the type of software used for the comparison. Some conclusions drawn belong to the performance, quality and accuracy of the above clustering algorithm.

Intelligent evaluation model of e-commerce transaction volume based on the combination of k-means and SOM algorithms

Intelligent evaluation model of e-commerce transaction volume based on the combination of k-means and SOM algorithms

Cancel-for-Any-Reason Insurance Recommendation Using Customer Transaction-Based Clustering

In the travel insurance industry, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">cancel-for-any-reason</i> insurance, also known as a cancellation protection service (CPS), is a recent attempt to strike a balance between customer satisfaction and service provider (SP) profits. However, some exceptional circumstances, particularly the COVID-19 pandemic, have led to a dramatic decrease in SP revenues, especially for non-refundable tickets purchased early with CPS. This paper begins by presenting a risk group segmentation of customers in an online ticket reservation system. Then, a CPS fee is recommended depending on the different customer risk groups provided by the cluster segmentation via different clustering algorithms such as centroid-based K-means, hierarchical agglomerative, DBSCAN, and artificial neural network-based SOM algorithms. According to the implemented cluster metrics, which include the Silhouette index, Davies-Bouldin index, Entropy index, and DBCV index, the SOM algorithm presents the most appropriate result. After predicting the new customer cluster, a CPS fee will be calculated with the proposed adaptive CPS method based on the cluster segmentation weights. Determining the weight of each cluster is related to the total CPS revenue threshold for all clusters defined by the SP. Therefore, to avoid a loss for SPs, the total CPS revenue will be kept constant with the threshold that the SP has been adjusted. The experimental results based on real-world data show that the risk group segmentation of customers helps to maintain a balance between CPS fees and SP profits. Finally, according to the calculated weights, the proposed model pegs the SP gain/loss variation with a 0.00012 exchange ratio.

Intelligent evaluation model of e-commerce transaction volume based on the combination of k-means and SOM algorithms

Due to the lack of steps to process sensitive data in the traditional intelligent evaluation model of e-commerce transaction volume, which leads to poor evaluation effect, an intelligent evaluation model of e-commerce transaction volume based on the combination of k-mean and SOM algorithm is proposed. Taking system intelligent clustering as the core, the transaction volume data in the web log is collected through the collection system. The desensitisation rule is used to establish the data transmission model of trading volume. The detection of sensitive data adopts signal detection technology and analytical processing. Based on k-means algorithm, data desensitisation is accomplished to achieve data denoising and a clustering evaluation model is established to complete data mining and analysis. The experimental results show that the model has a good denoising effect, and the ROC curve is closer to the upper left corner with an AUC value of 0.9712, which verifies the effectiveness and superiority of the model.

Identifying loyal customers and predicting customers purchase behavior using k-means and SOM algorithms

Despite the importance of data mining techniques to customer relationship management (CRM) and measuring customers loyalty and profitability, there is a lack of resources and articles related to this topic. Data mining is a useful tool to help companies for mining patterns and discovering hidden information in customers' data. In this study we cluster customers using k-means and SOM clustering algorithms with respect to apply RFM analysis based on behavioral characteristics such as recency, frequency and monetary variables and identify loyal customers and determine degree of loyalty. Then we apply C5.0 model on the resulting clusters to predict future customer behavior. In the end, evaluate accuracy of classification and compare the results. Proposed model implemented on M&S clothing store's dataset. Results of this study provide a background for identifying valuable and key customers and analysis their characteristics and loyalty.

The implementation of k-means partitioning algorithm in HOPACH clustering method

Hierarchical Ordered Partitioning And Collapsing Hybrid (HOPACH) is one of the powerful clustering methods which combine the strengths of partitioning and agglomerative clustering methods. Several partition clustering methods such as PAM, K-Means, SOM, or other partitioning algorithms can be used in the partitioning process. This process is followed by the ordering steps, then continued with the agglomerative process. The number of main clusters is determined by MSS (Mean Split Silhouette) value. MSS is used to measure the heterogeneity of the clustering result. The lower the MSS value, the more homogenous each cluster members. We select the number of clusters from the clustering results with minimum MSS. In this implementation of HOPACH, we incorporate k-Means partitioning algorithm in this HOPACH clustering method, to cluster and analyze 136 DNA sequences of Ebola viruses. The clustering process is started with collecting DNA sequences of Ebola viruses from GenBank, then followed by performing features extraction of these DNA sequences using N-Mers frequency. The extraction results are compiled to be a features matrix and normalized using the min-max normalization with the interval [0, 1] as an input data to generate genetic distance matrix using Euclidian distance. The genetic distance matrix is used in partitioning process by the K-Means algorithm in HOPACH clustering. As the results, we obtained 8 clusters with minimum MSS (Mean Split Silhouette) 0.50266. The clustering process in this article uses the open source program R.

Law, discipline and governance based on cognitive process simulation – analysis of legal forms in the modern rights relationship

People are becoming more and more humanized in the process of understanding the law. According to the right to discipline, the law has its own core setting factors, while some limits can't reach people's desire. Therefore, the legal and illegal mode of transcending rights is very important. In order to analyze the legal form of modern rights, in this paper, the cognitive learning and memory process of human brain were simulated through the artificial neural network and the understanding of human brain structure, and the role of law, discipline and governance was reflected. In the study, the structure and algorithm of the model neural network were optimized, the memory forgetting curve mechanism that can simulate the human brain was introduced, and thus the network recognition rate was improved. And in the algorithm, the calculation of matching degree was avoided, and the computational complexity was reduced to the sample. Then the sample was compared with the SOM, ART1, and PNN algorithms. The experimental simulation results show that the recognition speed of this sample is 1.9 times faster than that of ART1, 58 times than that of SOM, and 1.5 times than that of the PNN network.

A method of eliminating false alarm based on deep learning

Alarm data in intrusion detection system is mixed with a large amount of false alarm data, which brings great interference for network managers to analyze attack behavior. For a large deal of false alarm data in intrusion detection, this paper proposed an DBN construction method based on genetic operator improving the particle swarm, and used this DBN as a false alarm elimination classifier in IDS, firstly, using the improved particle swarm algorithm to search for the candidate network structure of DBN based on the fitness evaluation criteria, considering the candidate network structure with the optimal fitness as the final DBN network structure, secondly, using this DBN for false alarm elimination in intrusion detection. The experimental results showed that the average elimination rate of the proposed method is 5.54% and 2.9% higher than that of the SOM and KNN algorithms respectively, and the average misuse rate is 3.99% and 1.22% lower than that of the SOM and KNN algorithms respectively.

The use of ants algorithm in the detection and classification of intrusion in computer networks

The widespread development of using networks and relying on it and the emergence of many challenges faced by so many institutes like ministry of defense, intelligence agencies , banks, universitie...etc which depend heavily on service provided by those networks, resulted in the emergence of a new security concept towards building new security systems known as " Intrusion Detection Systems". The current study aims at detecting the known and unknown attacks as well as classifying the known ones. A network intrusion detection system is developed using AntMiner algorithm, which is one of the Swarm Intelligence which assimilate , the behavior of natural environment of ants. It is used to investigate the training data and getting the basic classification which is helpful in classifying the attacks in the tests data. The KDD Cup 1999 dataset concerned in constructing and evaluating intrusion systems which includes Training Dataset of 494021 in number as well as the testing dataset of 311029 in number. Experiments are conducted on the system using data sets and the complete tests. The results show that the detection rate using the SOM is: 92.37% and false alarm is: 4.67% and classification rate using Ant-Miner algorithm is : DOS(Denial of Service) 96.15 %, U2R(User to Root) 97.13%, Normal 94.50%, probe 72.90%, R2L(Remote to Local) 13.88%. These results show the effectiveness of SOM and Ant-Miner algorithms in the field of network security

A Performance Neighborhood Distance (ndist) Between K -Means and SOM Algorithms

A Performance Neighborhood Distance (<i>ndist</i>) Between <i>K</i> -Means and SOM Algorithms

The Classification of Diabetes Mellitus Using Kernel k-means

Diabetes Mellitus is a metabolic disorder which is characterized by chronicle hypertensive glucose. Automatics detection of diabetes mellitus is still challenging. This study detected diabetes mellitus by using kernel k-Means algorithm. Kernel k-means is an algorithm which was developed from k-means algorithm. Kernel k-means used kernel learning that is able to handle non linear separable data; where it differs with a common k-means. The performance of kernel k-means in detecting diabetes mellitus is also compared with SOM algorithms. The experiment result shows that kernel k-means has good performance and a way much better than SOM.

An algorithm J-SC of detecting communities in complex networks

Currently, community detection in complex networks has become a hot-button topic. In this paper, based on the Spectral Clustering (SC) algorithm, we introduce the idea of Jacobi iteration, and then propose a novel algorithm J-SC for community detection in complex networks. Furthermore, the accuracy and efficiency of this algorithm are tested by some representative real-world networks and several computer-generated networks. The experimental results indicate that the J-SC algorithm can accurately and effectively detect the community structure in these networks. Meanwhile, compared with the state-of-the-art community detecting algorithms SC, SOM, K-means, Walktrap and Fastgreedy, the J-SC algorithm has better performance, reflecting that this new algorithm can acquire higher values of modularity and NMI. Moreover, this new algorithm has faster running time than SOM and Walktrap algorithms.