Alok Kumar Jagadev

Multi-objective Swarm Intelligence

The aim of this book is to understand the state-of-the-art theoretical and practical advances of swarm intelligence. It comprises seven contemporary relevant chapters. In chapter 1, a review of Bacteria Foraging Optimization (BFO) techniques for both single and multiple criterions problem is presented. A survey on swarm intelligence for multiple and many objectives optimization is presented in chapter 2 along with a topical study on EEG signal analysis. Without compromising the extensive simulation study, a comparative study of variants of MOPSO is provided in chapter 3. Intractable problems like subset and job scheduling problems are discussed in chapters 4 and 7 by different hybrid swarm intelligence techniques. An attempt to study image enhancement by ant colony optimization is made in chapter 5. Finally, chapter 7 covers the aspect of uncertainty in data by hybrid PSO.

Honey Bee Behavior: A Multi-agent Approach for Multiple Campaigns Assignment Problem

This paper address a multi-agent approach using the behavior of honey bee to find out an optimal customer-campaign relationship under certain restrictions for the problem of multiple campaigns assignment. This NP-hard problem is one of the key issues in marketing when producing the optimal campaign. In personalized marketing it is very important to optimize the customer satisfaction and targeting efficiency. Using the behavior of honey bee a multi-agent approach is proposed to overcome the multiple recommendations problem that occur when several personalized campaigns conducting simultaneously. We measure the effectiveness of the propose method with two other methods known as RANDOM and INDEPENDENT using an artificially created customer-campaign preference matrix. Further a generalized Gaussian response suppression function is introduced and it differs among customer classes. An extensive simulation studies are carried out varying on the small to large scale of the customer-campaign assignment matrix and the percentage of recommendations. Computational result of the proposed method shows a clear edge vis-a-vis RANDOM and INDEPENDENT.

Modified Ant Colony Optimization Algorithm (MAnt-Miner) for Classification Rule Mining

Classification rule mining is an important task of data mining. Ant colony optimization (ACO) algorithms are applied successfully to various optimization problems. Earlier Ant-Miner, an ACO algorithm was used to discover the classification rules and predictive accuracy was determined. In this paper, modified ant colony optimization (MAnt-Miner) is proposed to generate the classification rules and to enhance the predictive accuracy. This method is applied on breast cancer data set, and the experimental result showed that the predictive accuracy of MAnt-Miner is better than Ant-Miner.

Optimized Radial Basis Functional neural network for stock index prediction

Prediction of stock market indices is an interesting and challenging research problem in financial data mining area because movement of stock indices are nonlinear and they are dependent upon different constitutional and extraneous aspects. In this paper we come up with the practice of different techniques of Artificial Neural Network (ANN) in stock market prediction. Here we have selected Multilayer Perceptron model (MLP), Radial Basis Functional Network (RBFN) and an optimized Radial Basis Functional neural network. The proposed model provides an optimized set of center set to Radial Basis Functional Network for experiment and Particle swarm algorithm is used for this boosting process. Lastly in order to verify the best tactics, a comparative result of the applied models is represented.

An Empirical Analysis of Training Algorithms of Neural Networks: A Case Study of EEG Signal Classification Using Java Framework

With the pace of modern lifestyle, about 40–50 million people in the world suffer from epilepsy—a disease with neurological disorder. Electroencephalography (EEG) is the process of recording brain signals that generate due to a small amount of electric discharge in brain. This may occur due to the information flow among several neurons. Therefore, in every minute, analysis of EEG signal can solve much neurological disorders like epilepsy. In this paper, a systematic procedure for analysis and classification of EEG signal is discussed for identification of epilepsy in a human brain. The analysis of EEG signal is made through a series of steps from feature extraction to classification. Feature extraction from EEG signal is done through discrete wavelet transform (DWT), and the classification task is carried out by MLPNN based on supervised training algorithms such as backpropagation, resilient propagation (RPROP), and Manhattan update rule. Experimental study in a Java platform confirms that RPROP trained MLPNN to classify EEG signal is promising as compared to back-propagation or Manhattan update rule trained MLPNN.

Forecasting Methods in Electric Power Sector

Electricpowerplaysavibrantroleineconomicgrowthanddevelopmentofaregion.Thereisastrong co-relationbetweenthehumandevelopmentindexandpercapitaelectricityconsumption.Providing adequateenergyofdesiredqualityinvariousformsinasustainablemannerandatacompetitiveprice isoneofthebiggestchallenges.Tomeetthefast-growingelectricpowerdemand,onasustained basis,meticulouspowersystemplanningisrequired.Thisplanningneedselectricalloadforecasting asitprovidestheprimaryinputsandenablesfinancialanalysis.Accurateelectricloadforecastsare helpfulinformulatingloadmanagementstrategiesinviewofdifferentemergingeconomicscenarios, whichcanbedovetailedwiththedevelopmentplanoftheregion.Theobjectiveofthisarticleisto understandvariouslongtermelectricalloadforecastingtechniques,toassessitsapplicability;and usefulnessfor longtermelectrical loadforecastingforanisolatedremoteregion,underdifferent growthscenariosconsideringdemandsidemanagement,priceandincomeeffect. KEywORdS Artificial Neural Network, Electrical Energy Consumption, Electrical Energy Requirements, Long Term Electrical Load Forecasting, Parametric

A cross mutation-based differential evolution for data clustering

A cross mutation-based differential evolution (CMDE) approach is proposed here to handle the complexity issue in clustering due to the data uncertainty, overlapping and rapid growth in size of data. In this CMDE, a novel mutation strategy and a centroid rearrangement scheme have been proposed for getting a better and consistent result. CMDE provides optimal cluster centres with minimum intra cluster distance and maximum accuracy percentage. A comparative analysis of the proposed approach with another five population based methods, such as dynamic shuffled differential evolution (DSDE), ant colony optimisation (ACO), artificial bee colony (ABC), particle swarm optimisation (PSO) and particle swarm optimisation with age-group topology (PSOAG) is carried out to justify the better clustering performance of the suggested method. These techniques are applied to seven real datasets and the results reveal the efficacy of the proposed algorithm for clustering in various fields.

Comparison of Various Approaches in Multi-objective Particle Swarm Optimization (MOPSO): Empirical Study

This chapter presents a study of particle swarm optimization (PSO) method in multi-objective optimization problems. Many of these methods have focused on improving characteristics like convergence, diversity, and computational times by proposing effective ‘archiving’ and ‘guide selection’ techniques. What has still been lacking is an empirical study of these proposals in a common frame-work. In this chapter, an attempt to analyze these methods has been made; discussing their strengths and weaknesses. A multi-objective particle swarm optimization (MOPSO) algorithm, named dynamic multiple swarms in MOPSO is compared with other well known MOPSO techniques in which the number of swarms are adaptively adjusted throughout the search process via dynamic swarm strategy. The strategy allocates an appropriate number of swarms as required to support convergence and diversity criteria among the swarms. Additional novel designs include a PSO updating mechanism to better manage the communication within a swarm and among swarms and an objective space compression and expansion strategy to progressively exploit the objective space during the search process. Comparative study shows that the performance of the variant is competitive in comparison to the selected algorithms on standard benchmark problems. A dynamic MOPSO approach is compared and validated using several test functions and metrics taken from the standard literatures on evolutionary multi-objective optimization. Results indicate that the approach is highly competitive and that can be considered a viable alternative to solve multi-objective optimization problems.

Learning an Artificial Neural Network Using Dynamic Particle Swarm Optimization–Backpropagation: Empirical Evaluation and Comparison

Training neural networks is a complex task with great importance in the field of supervised learning. In the training process, a set of input?output patterns is repeated to an artificial neural network (ANN). From those patterns weights of all the interconnections between neurons are adjusted until the specified input yields the desired output. In this paper, a new hybrid algorithm is proposed for global optimization of connection weights in an ANN. Dynamic swarms are shown to converge rapidly during the initial stages of a global search, but around the global optimum, the search process becomes very slow. In contrast, the gradient descent method can achieve faster convergence speed around the global optimum, and at the same time, the convergence accuracy can be relatively high. Therefore, the proposed hybrid algorithm combines the dynamic particle swarm optimization (DPSO) algorithm with the backpropagation (BP) algorithm, also referred to as the DPSO-BP algorithm, to train the weights of an ANN. In this paper, we intend to show the superiority (time performance and quality of solution) of the proposed hybrid algorithm (DPSO-BP) over other more standard algorithms in neural network training. The algorithms are compared using two different datasets, and the results are simulated.

Machine learning-based classifiers ensemble for credit risk assessment

Credit risk assessment is acting as a survival weapon in almost every financial institution. It involves deep and sensitive analysis of various financial, social, demographic and other pertinent data provided by the customers and about the customers for building a more accurate and robust electronic finance system. The classification problem is one of the major concerned in the process of analysing gamut of data; however, its complexity has ignited us to use machine learning-based approaches. In this paper, some machine learning algorithms have been studied and compared their effectiveness for credit risk assessment. Further, as an extension of our study, we develop a novel sliding window-based meta-majority voting ensemble learning to improve the prediction accuracy of credit risk assessment problem by properly analysing the underlying samples. The experimental findings draw a clear line between the proposed ensembler and traditional ensemblers. Moreover, the proposed method is very promising vis-a-vis of individual classifiers.