LVQ Network Research Articles

Audit credit guarantee risk identification method based on LM algorithm

Aiming at the problem of low recognition accuracy caused by the high probability of training model falling into local extreme value in audit credit guarantee risk identification, an audit credit guarantees risk identification method based on LM algorithm is proposed. The risk identification index system is established and its weight is assigned to provide the index basis of the input layer for the identification model. The BP neural network is improved by LM algorithm to reduce the probability of falling into local extreme value, and the audit credit guarantee risk identification model is established by using the improved network. Using the same audit credit guarantee risk sample to test the identification method, the identification accuracy of this method is 95.15%, which is 4.86% and 4.61% higher than that based on BP neural network and LVQ network respectively. Therefore, the identification results of the design method are more feasible.

Analisis Perbandingan Metode LVQ Dan Backpropagation dalam Penentuan Keaslian Uang Kertas Rupiah Berbasis Parameter HSV

The high demand of the community for money that causes crime is the circulation of fake paper money. Counterfeit money if shared has the same physical as the original money issued by Bank Indonesia. To avoid the public accidentally transacting using counterfeit money, the government has actually socialized the 3D method (Seen, Diraba, and Diterawang). However, along with technological developments, the technique of making counterfeit money will also require the development of alternative techniques that can be used to help save fake money. Determining the authenticity of Rupiah banknotes can be done using pattern classification methods, one of which can be accommodated by artificial neural networks. LVQ neural network (Learning Vector Quantization) and Backpropagation are two types of artificial neural networks that do supervised learning. Extraction of features that show the authenticity of banknotes can be done using the HSV color space. This color space consists of components H (Hue), S (Saturation), and V (Value). This is the background of the topic chosen for analysis choosing the LVQ and Backpropagation methods in determining the authenticity of Rupiah banknotes based on HSV parameters. Evaluation analysis taken from the level assessment The evaluation results using the composition of the test data consisting of 10 original money images and 8 original money images obtained results both LVQ and Backpropagation networks were able to classify real and fake money images with 100% acquisition rates. However, when viewed from the MSE value, the LVQ network has a better performance with the supporting MSE value being 0. The test results from the preparation time, the Backpropagation network requires a shorter time compared to the LVQ network.

Neural based RBF and LVQ Network Model of Knowledge Representation in the Prediction of Mobile Location

Neural based RBF and LVQ Network Model of Knowledge Representation in the Prediction of Mobile Location

Simplifying credit scoring rules using LVQ + PSO

PurposeOne of the key elements in the banking industry relies on the appropriate selection of customers. To manage credit risk, banks dedicate special efforts to classify customers according to their risk. The usual decision-making process consists of gathering personal and financial information about the borrower. Processing this information can be time-consuming, and presents some difficulties because of the heterogeneous structure of data.Design/methodology/approachThis paper presents an alternative method that is able to generate rules that work not only on numerical attributes but also on nominal ones. The key feature of this method, called learning vector quantization and particle swarm optimization (LVQ + PSO), is the finding of a reduced set of classifying rules. This is possible because of the combination of a competitive neural network with an optimization technique.FindingsThese rules constitute a predictive model for credit risk approval. The reduced quantity of rules makes this method useful for credit officers aiming to make decisions about granting a credit. It also could act as an orientation for borrower’s self evaluation about her/his creditworthiness.Research limitations/implicationsIn spite of the fact that conducted tests showed no evidence of dependence between results and the initial size of the LVQ network, it is considered desirable to repeat the measurements using an LVQ network of minimum size and a version of variable population PSO to adequately explore the solution space in the future.Practical implicationsIn the past decades, there has been an increase in consumer credit. Retail banking is a growing industry. Not only has there been a boom in credit card memberships, specially in emerging economies, but also an increase in small consumption credits. For example, it is very common in emerging economies that families buy home appliances on installments. In those countries, the association of a home appliance shop with a financial institution is usual, to provide customers with quick-decision credit line facilities. The existence of such a financial instrument aids to boost sales. This association generates conflict of interests. On one hand, the home appliance shop wants to sell products to all customers. Therefore, it is in its best interest to promote a generous credit policy. On the other hand, the financial institution wants to maximize the revenue from credits, leading to a strict surveillance of loan losses. Having a fair and transparent credit-granting policy favors a good business relationship between home appliances shops and financial institutions. One way of developing such a policy is to construct objective rules to decide to grant or deny a credit application.Social implicationsBetter credit decision rules generate enhanced risk sharing. In addition, it improves transparency in credit acceptance decisions, giving less room to arbitrary decisions.Originality/valueThis study develops a new method that combines a competitive neural network and an optimization technique. It was applied to a real database of a financial institution in a developing country.

PERBANDINGAN TINGKAT AKURASI JENIS CITRA KEABUAN , HSV, DAN L*a*b* PADA IDENTIFIKASI JENIS BUAH PIR

Image processing has been commonly used in automatic object identification. These are some methods that can be used for automatic object identification, such as LVQ, K-NN, SVM, and Neural Network. This research specifically bring out the topic about the level accuracy comparison in identification of pear variety using grayscale, HSV, and L*a*b* images in aim to get the best image type for pear image identification using neural network. The feature are gray level co-occurrence matrix feature (energy, entropy, homogeneity, and contrast) from canny edge detection’s image and also color feature. Based on image examination result, grayscale reached its best accuracy for 90% on MSE 1e-10 with 10 hidden layer neurons, HSV reached its best accuracy for 100% on MSE 1e-5 with 20 hidden layer neurons, L*a*b* reached its best accuracy for 100% on MSE 1e-5 with 15 hidden layer neurons. HSV and L*a*b* give the better accuracy for pear variety image identification than grayscale.Keyword:Image Processing, Pear, Neural Network, Identification, Gray Level Co-occurrence Matrix, Canny, Color.

An optimization Algorithm of LVQ Neural Network Classification by GA

This paper describes a dimension reduction method of input vector to improve classification efficiency of LVQ neural network, where GA is used to decrease the redundancy of input data. And in order to solve the initial weight vector sensitivity, GA is also employed to optimize the initial vector. The experimental results on the UCI data sets demonstrate that the efficiency and accuracy of our LVQ network by GA is higher than general LVQ neural network classification algorithm.

Vision Location Algorithm of Circular Hole in Distributed Clutter Background Using LVQ Network

Vision Location Algorithm of Circular Hole in Distributed Clutter Background Using LVQ Network

Comparing Frei-Chen & Sobel Edge Detector in Face Emotion Recognition with Euclidean Distance and LVQ Network

In this paper, to increase the execution speed in face emotion recognition instead Sobel edge detector of Frei-Chen edge detector is used. The advantages Frei-Chen method are accurate and better detection, low overhead, detect corners beside with edges and less sensitive to noise than other edge detector. The Frei-Chen algorithm works on a part of image with size 3*3 similarly like the Sobel filter but instead of two masks of 4 masks are used. Of important characteristics this filter can be pointed to 4 masks of unique. For edge detection suitable masks are selected and selected mask is applied on image. This filter such as sobel filter can be used for multi-channel images. Furthermore we know Facial expression gives important information about emotion of a person. Face emotion recognition is one of important issues that widely attended in recent years. It can be used in areas of security, control, entertainment and machine vision. Nowadays for emotion recognition use of science image processing, speech signal processing, gesture signal processing and physiological signal processing. Use of image processing science than other methods can be very useful and helpful. Our proposal uses of an objective function for minimization of sum of Euclidean distance from the given points to the eye and lip curve that PSO algorithm will be used to optimize objective function. We recommend use of ellipse form as eye and lip in face emotion recognition. Face emotion recognition process like presented previous papers involves three stages pre-processing, feature extraction and classification. One of biggest problems in classification of emotions is overlap in range of values. To increase success rate in face emotion recognition we in this paper for another experience of neural networks use of LVQ neural network. We also show experiments face emotion recognition with Sobel filter and Frei-Chen filter. The results show that using Frei-Chen filter against Sobel filter reduces run time. Keywords: Projection Profile, Particle Swarm Optimization (PSO), Euclidean Distance Measurement LVQ Network and Frei-Chen Edge Detector.

Premature clustering phenomenon and new training algorithms for LVQ

Five existing LVQ algorithms are reviewed. The Premature Clustering Phenomenon, which downgrades the performance of LVQ is explained. By introducing and applying the “ equalizing factor” as a remedy for the premature clustering phenomenon a breakthrough is achieved in improving the performance of the LVQ network, and its performance becomes competitive with that of the best known classifiers. For estimating the equalizing factor four different formulas are suggested, which result in four different versions of the LVQ4a algorithm. A new weight-updating formula for LVQ is presented, and the LVQ4b algorithm is presented as implementation of this new weight-updating formula in batch mode training. In addition, four variants of the LVQ4c algorithm are presented as the customized LVQ4b algorithm for pattern mode training. A meticulous analysis of their performances and that of five early training algorithms has been carried out and they have been compared against each other, on 16 databases of the Farsi optical character recognition problem.

ADAPTIVE CLASSIFIER INTEGRATION FOR INVARIANT FACE RECOGNITION (ACIIFR)

The integration of multiple classifiers promises higher classification accuracy and robustness than can be obtained with a single classifier. We address two problems: (a) automatic recognition of human faces using a novel fusion approach based on an adaptive LVQ network architecture, and (b) improve the face recognition up to 100% while maintaining the learning time per face image constant, which is an scalability issue. The learning time per face image of the recognition system remains constant irrespective of the data size. The integration of the system incorporates the "divide and conquer" modularity principles, i.e. divide the learning data into small modules, train individual modules separately using compact LVQ model structure and still encompass all the variance, and fuse trained modules to achieve recognition rate nearly 100%. The concept of Merged Classes (MCs) is introduced to enhance the accuracy rate. The proposed integrated architecture has shown its feasibility using a collection of 1130 face images of 158 subjects from three standard databases, ORL, PICS and KU. Empirical results yield an accuracy rate of 100% on the face recognition task for 40 subjects in 0.056 seconds per image. Thus, the system has shown potential to be adopted for real time application domains.

Data Analysis of Debris Flow with Soft Computing

The purpose of this study is to develop the analysis method for debris flow with soft computing. The soft computing methods which embrace neural networks and/or genetic algorithms (GA) are expected to solve problems which are defined imprecisely from a complex phenomena.The neural network combined with the GA was introduced into the analysis for the occurrence condition of debris flow. Training data sets to the LVQ network are optimized by the GA. Theprocedure was used to examine a characteristic change in occurrence condition of debris flow. Runoff analysis of debris flow by the use of GMDH was proposed. These methods were applied to the data for the Mizunashi River in the Unzen Volcano. The results led that the soft computing methods are useful for data analysis of the debris flow.

Near Infrared Spectroscopy for Classification of Iberian Pig Carcasses Using an Artificial Neural Network

Artificial neural networks (ANNs) have demonstrated their usefulness in near infrared (NIR) reflection and transmittance spectroscopy for quantitative prediction. The new approach presented here considers the use of ANNs for qualitative classification. Four forms of neural networks (a competitive network using the learning vector quantisation, LVQ learning rule; a backpropagation network using the extended delta-bar-delta, EDBD rule; a network with direct random search, DRS; and a simple competitive linear network, CL) have been tested for classification of 118 fat samples from Iberian pig carcasses into three different price groups. An ANN using the LVQ learning rule has been found to be the best in terms of classification error size. The classification ability of the LVQ network has been evaluated against discriminant analysis, one of the most used methods for NIR spectroscopic qualitative analysis.