Janmenjoy Nayak

Fuzzy C-Means (FCM) Clustering Algorithm: A Decade Review from 2000 to 2014

The Fuzzy c-means is one of the most popular ongoing area of research among all types of researchers including Computer science, Mathematics and other areas of engineering, as well as all areas of optimization practices. Several problems from various areas have been effectively solved by using FCM and its different variants. But, for efficient use of the algorithm in various diversified applications, some modifications or hybridization with other algorithms are needed. A comprehensive survey on FCM and its applications in more than one decade has been carried out in this paper to show the efficiency and applicability in a mixture of domains. Also, another intention of this survey is to encourage new researchers to make use of this simple algorithm (which is popularly called soft classification model) in problem solving.

A self adaptive harmony search based functional link higher order ANN for non-linear data classification

In the data classification process involving higher order ANNs, its a herculean task to determine the optimal ANN classification model due to non-linear nature of real world datasets. To add to the woe, it is tedious to adjust the set of weights of ANNs by using appropriate learning algorithm to obtain better classification accuracy. In this paper, an improved variant of harmony search (HS), called self-adaptive harmony search (SAHS) along with gradient descent learning is used with functional link artificial neural network (FLANN) for the task of classification in data mining. Using its past experiences, SAHS adjusts the harmonies according to the maximum and minimum values in the current harmony memory. The powerful combination of this unique strategy of SAHS and searching capabilities of gradient descent search is used to obtain optimal set of weights for FLANN. The proposed method (SAHS-FLANN) is implemented in MATLAB and the results are contrasted with other alternatives (FLANN, GA based FLANN, PSO based FLANN, HS based FLANN, improved HS based FLANN and TLBO based FLANN). To illustrate its effectiveness, SAHS-FLANN is tested on various benchmark datasets from UCI machine learning repository by using 5-fold cross validation technique. Under the null-hypothesis, the proposed method is analyzed by using various statistical tests for statistical correctness of results. The performance of the SAHS-FLANN is found to be better and statistically significant in comparison with other alternatives. The SAHS-FLANN differs from HS-FLANN (HS based FLANN) by the elimination of constant parameters (bandwidth and pitch adjustment rate). Furthermore, it leads to the simplification of steps for the improvisation of weight-sets in IHS-FLANN (improved HS based FLANN) by incorporating adjustments of new weight-sets according to the weight-sets with maximum and minimum fitness. We have proposed a novel approach of hybridization of higher order neural network (Functional link higher order artificial neural network) with self adaptive harmony search (SAHS) based gradient descent learning (GDL) for non-linear data classification problem.Proposed approach exhibits better performance than other alternative approaches.Statistical analysis has been performed by using various statistical methods (Friedman test, Holm and Hochberg procedure, Tukey test and Dunnett test) under null-hypothesis in order to prove the proposed method is statistically valid and better.

Hybrid Clustering using Elitist Teaching Learning-Based Optimization: An Improved Hybrid Approach of TLBO

Data clustering is a key field of research in the pattern recognition arena. Although clustering is an unsupervised learning technique, numerous efforts have been made in both hard and soft clustering. In hard clustering, K-means is the most popular method and is being used in diversified application areas. In this paper, an effort has been made with a recently developed population based metaheuristic called Elitist based teaching learning based optimization ETLBO for data clustering. The ETLBO has been hybridized with K-means algorithm ETLBO-K-means to get the optimal cluster centers and effective fitness values. The performance of the proposed method has been compared with other techniques by considering standard benchmark real life datasets as well as some synthetic datasets. Simulation and comparison results demonstrate the effectiveness and efficiency of the proposed method.

A Novel FLANN with a Hybrid PSO and GA Based Gradient Descent Learning for Classification

In this paper, it is an attempt to design a PSO & GA based FLANN model (PSO-GA-FLANN) for classification with a hybrid Gradient Descent Learning (GDL). The PSO, GA and the gradient descent search are used iteratively to adjust the parameters of FLANN until the error is less than the required value. Accuracy and convergence of PSO-GA-FLANN is investigated and compared with FLANN, GA-based FLANN and PSO-based FLANN. These models have been implemented and results are statistically analyzed using ANOVA test in order to get significant result. To obtain generalized performance, the proposed method has been tested under 5-fold cross validation.

A Harmony Search Based Gradient Descent Learning-FLANN (HS-GDL-FLANN) for Classification

The Harmony Search (HS) algorithm is meta-heuristic optimization inspired by natural phenomena called musical process and it quite simple due to few mathematical requirements and simple steps as compared to earlier meta-heuristic optimization algorithms. It mimics the local and global search procedure of pitch adjustment during production of pleasant harmony by musicians. Although HS has been used in many application like vehicle routing problems, robotics, power and energy etc., in this paper, an attempt is made to design a hybrid FLANN with Harmony Search based Gradient Descent Learning for classification. The proposed algorithm has been compared with FLANN, GA based FLANN and PSO based FLANN classifier to get remarkable performance. All the four classifier are implemented in MATLAB and tested by couples of benchmark datasets from UCI machine learning repository. Finally, to get generalized performance, 5 fold cross validation is adopted and result are analyzed under one-way ANOVA test.

Particle Swarm Optimization Based Higher Order Neural Network for Classification

The maturity in the use of both the feed forward neural network and Multilayer perception brought the limitations of neural network like linear threshold unit and multi-layering in various applications. Hence, a higher order network can be useful to perform nonlinear mapping using the single layer of input units for overcoming the drawbacks of the above-mentioned neural networks. In this paper, a higher order neural network called Pi-Sigma neural network with standard back propagation Gradient descent learning and Particle Swarm Optimization algorithms has been coupled to develop an efficient robust hybrid training algorithm with the local and global searching capabilities for classification task. To demonstrate the capacity of the proposed PSO-PSNN model, the performance has been tested with various benchmark datasets from UCI machine learning repository and compared with the resulting performance of PSNN, GA-PSNN. Comparison result shows that the proposed model obtains a promising performance for classification problems.

A hybrid PSO-GA based Pi sigma neural network (PSNN) with standard back propagation gradient descent learning for classification

Due to the strong global optimization capability and fast convergence, PSO has shown its efficiency in solving various real world benchmark applications. But premature convergence is one of the major drawback of PSO. In this paper to address this issue, a hybrid PSO-GA based Pi-sigma neural network with standard back propagation gradient descent learning (PSO-GA-PSNN) has been proposed for classification problems. The adjustment of algorithmic parameters is iteratively used until the error is less than the desired output. The proposed PSO-GA-PSNN has been tested with various benchmark datasets taken from UCI machine learning repository and the simulated results are being tested with the statistical tool ANOVA to show the obtained results are statistically steady and valid.

An Improved Firefly Fuzzy C-Means (FAFCM) Algorithm for Clustering Real World Data Sets

Fuzzy c-means has been widely used in clustering many real world datasets used for decision making process. But sometimes Fuzzy c-means (FCM) algorithm generally gets trapped in the local optima and is highly sensitive to initialization. Firefly algorithm (FA) is a well known, popular metaheuristic algorithm that simulates through the flashing characteristics of fireflies and can be used to resolve the shortcomings of Fuzzy c-means algorithm. In this paper, first a firefly based fuzzy c-means clustering and then an improved firefly based fuzzy c-means algorithm (FAFCM) has been proposed and their performance are being compared with fuzzy c-means and PSO algorithm. The experimental results divulge that the proposed improved FAFCM method performs better and quite effective for clustering real world datasets than FAFCM, FCM and PSO, as it avoids to stuck in local optima and leads to faster convergence.

An efficient FLANN model with CRO-based gradient descent learning for classification

Due to the nonlinear nature of real world data, it is difficult to determine the optimal ANN classification model with accurate and fast convergence. Although, many higher order ANN have been designed and integrated with competitive optimisation method in order to construct an accurate classification model, but the parameter adjustment and variability in performance in different runs of the classification model leads to statistically insignificant result. In this paper, a FLANN model CRO-GDL-FLANN has been proposed for classification with gradient descent learning GDL based on chemical reaction optimisation CRO. The proposed CRO-GDL-FLANN method has been tested with various benchmark datasets from the UCI machine learning repository under five fold cross-validations. The classification accuracy of CRO-GDL-FLANN is compared with FLANN, GA-FLANN and PSO-FLANN. To prove the proposed method is statistically better and significantly different from other alternatives, the CRO-GDL-FLANN is verified under multiple comparisons of classifiers by using Friedman, Tukey and Dunnett statistical test. Finally, one-way-ANOVA test has been carried out for generalised comparison of CRO-GDL-FLANN with other classifiers.

Evolutionary Improved Swarm-Based Hybrid K-Means Algorithm for Cluster Analysis

Improvement in the quality of cluster centers and minimization of intra-cluster distance are two most challenging areas of K-means clustering algorithm. Due to predetermined number of clusters, it is difficult to predict the exact value of k. Furthermore, in case of non-globular clusters, K-means fails to get optimal cluster center in a data set. In this paper, a hybrid improved particle swarm optimization-based evolutionary K-means clustering method has been proposed to obtain the optimal cluster center. The hybridization of improved PSO and genetic algorithm (GA) along with K-means algorithm improves the convergence speed as well as helps to find the global optimal solution. In the first stage, IPSO has been used to get a global solution in order to get optimal cluster centers. Then, the crossover steps of GA are used to improve the quality of particles and mutation is used for diversification of solution space in order to avoid premature convergence. The performance analysis of the proposed method is compared with other existing clustering techniques like K-means, GA-K-means, and PSO-K-means.