ICTACT Journal on Soft Computing

M. Balamurugan ¹, S. Kannan ¹

Affiliations
1 Department of Computer Applications, Madurai Kamaraj University, IN

Abstract

In Ensemble classifiers, the Combination of multiple prediction models of classifiers is important for making progress in a variety of difficult prediction problems. Ensemble of classifiers proved potential in getting higher accuracy compared to single classifier. Even though by the usage ensemble classifiers, still there is in-need to improve its performance. There are many possible ways available to increase the performance of ensemble classifiers. One of the ways is sampling, which plays a major role for improving the quality of ensemble classifier. Since, it helps in reducing the bias in input data set of ensemble. Sampling is the process of extracting the subset of samples from the original data set. In this research work, analysis is done on sampling techniques for ensemble classifiers. In ensemble classifier, specifically one of the probability based sampling techniques is being always used. Samples are gathered in a process which gives all the individuals in the population of equal chances, such that, sampling bias is removed. In this paper, analyse the performance of ensemble classifiers by using various sampling techniques and list out their drawbacks.

Keywords

Ensemble of Classifiers, Sampling, Random Forest, Boosting.

Full Text

     

S. Saravanan ¹, S. Kannan ², C. Thangaraj ³

Affiliations
1 Department of Electrical and Electronics Engineering, Kalasalingam University, IN
2 Department of Electrical and Electronics Engineering, Ramco Institute of Technology, IN
3 Vignan University, IN

Abstract

This study aims to provide an accurate and realistic prediction model for electricity demand using population, imports, exports, per capita Gross Domestic Product (GDP) and per capita Gross National Income (GNI) data for India. Four different models were used for different combinations of the above five input variables and the effect of input variables on the estimation of electricity demand has been demonstrated. In order to train the network 29 years data and to test the network 9 years data have been used. The future electricity demand for a period of 8 years from 2013 to 2020 has been predicted. The performance of the ANFIS technique is proved to be better than Multiple Linear Regression (MLR) and Artificial Neural Networks (ANN).

Keywords

ANFIS, ANN, Exports, GDP, GNI, Imports, Load Forecasting, MLR.

Full Text

     

K. Karthikeyan ¹, S. Kannan ¹, S. Baskar ², C. Thangaraj ³

Affiliations
1 Department of Electrical and Electronics Engineering, Kalasalingam University, IN
2 Department of Electrical and Electronics Engineering, Thiagarajar College of Engineering, IN
3 Anna University of Technology, Chennai, IN

Abstract

This paper describes the application of an evolutionary algorithm, Restart Covariance Matrix Adaptation Evolution Strategy (RCMAES) to the Generation Expansion Planning (GEP) problem. RCMAES is a class of continuous Evolutionary Algorithm (EA) derived from the concept of self-adaptation in evolution strategies, which adapts the covariance matrix of a multivariate normal search distribution. The original GEP problem is modified by incorporating Virtual Mapping Procedure (VMP). The GEP problem of a synthetic test systems for 6- year, 14-year and 24-year planning horizons having five types of candidate units is considered. Two different constraint-handling methods are incorporated and impact of each method has been compared. In addition, comparison and validation has also made with dynamic programming method.

Keywords

Constraint Handling, Dynamic Programming, Generation Expansion Planning, Restart Covariance Matrix Adaptation Evolution Strategy and Virtual Mapping Procedure.

Full Text

     

S. Saravanan ¹, S. Kannan ¹, C. Thangaraj ²

Affiliations
1 Department of Electrical and Electronics Engineering, Kalasalingam University, IN
2 Anna University of Technology, Chennai, IN

Abstract

Power System planning starts with Electric load (demand) forecasting. Accurate electricity load forecasting is one of the most important challenges in managing supply and demand of the electricity, since the electricity demand is volatile in nature; it cannot be stored and has to be consumed instantly. The aim of this study deals with electricity consumption in India, to forecast future projection of demand for a period of 19 years from 2012 to 2030. The eleven input variables used are Amount of CO₂ emission, Population, Per capita GDP, Per capita gross national income, Gross Domestic savings, Industry, Consumer price index, Wholesale price index, Imports, Exports and Per capita power consumption. A new methodology based on Artificial Neural Networks (ANNs) using principal components is also used. Data of 29 years used for training and data of 10 years used for testing the ANNs. Comparison made with multiple linear regression (based on original data and the principal components) and ANNs with original data as input variables. The results show that the use of ANNs with principal components (PC) is more effective.

Keywords

Artificial Neural Networks, Electricity Load Forecasting, Regression Analysis, Principal Components.

Full Text

     

Balamurugan Mahalingam ¹, S. Kannan ¹, Vairaprakash Gurusamy ¹

Affiliations
1 Department of Computer Applications, Madurai Kamaraj University, IN

Abstract

Ensemble of classifiers combines the more than one prediction models of classifiers into single model for classifying the new instances. Unbiased samples could help the ensemble classifiers to build the efficient prediction model. Existing sampling techniques fails to give the unbiased samples. To overcome this problem, the paper introduces a k-modes random sample technique which combines the k-modes cluster algorithm and simple random sampling technique to take the sample from the dataset. In this paper, the impact of random sampling technique in the Ensemble learning algorithm is shown. Random selection was done properly by using k-modes random sampling technique. Hence, sample will reflect the characteristics of entire dataset.

Keywords

Sampling, Ensemble Classifiers, Cluster Random Sample.

Full Text