D. Ashok Kumar ¹, N. Kannathasan ²

Affiliations
1 Department of Computer Science, Government Arts College, Tiruchirapalli - 620022, Tamil Nadu, IN
2 Department of Computer Science, Kanchi Mamunivar Centre for Post Graduate Studies, Puducherry - 605008, U.T. of Puducherry, IN

Abstract

Background/Objectives: Data Mining has been used to analyze large datasets and establish useful classification and patterns in the datasets. The efficient analysis of data in different format becomes a challenging work. Methods/Statistical Analysis: This work proposed a novel Soil Profile Feature Reduction Model using Principal Component Analysis for data reduction. The proposed model uses the method of k-Means clustering and PC Aapproach for feature reduction which initially applies PCA to acquire reduced uncorrelated attributes showing maximal eigenvalues in the dataset with minimum loss of information. Again proposed model uses k-Means on the PCA reduced dataset to find out discriminative features that will be the most sufficient ones for classification. Findings: The weight by PCA generates attribute weights of the soil profile dataset using a component created by the PCA. The component is mentioned by the component number parameter. The normalize weight parameter is usually set to true to spread the weights between 0 and 1. The attribute weights reflect the relevance of the attributes with respect to the class attribute. The higher weight of an attribute is more relevant, it is considered. This is a combination of clustering approach with feature reduction to get a minimal set attributes relating a suitably high accuracy in describing the original features. The result of clustering is same after reducing the attributes using PCA. The experimental results prove that proposed model is reducing number of initial attributes, reducing computational complexity and improving predictive accuracy in High Dimensional Datasets. Applications/Improvements: The same soil profile feature is implemented by using the other techniques instead of PCA algorithm in future.

Keywords

Clustering, Feature Reduction, k-Means,Principal Component Analysis, Soil Profile

Full Text

   

D. Ashok Kumar ¹, N. Kannathasan ²

Affiliations
1 Department of Computer Science, Government Arts College, Tiruchirapalli - 620022, Tamil Nadu, IN
2 Department of Computer Science, Kanchi Mamunivar Centre for Post Graduate Studies Puducherry – 605008, U.T. of Puducherry, IN

Abstract

Background/Objectives: Fuzzy Logic is derived from Fuzzy Set that deals with reasoning that is accurate rather than precisely based on degrees of membership, and are well-liked tools in the application of prediction, classification and recognition based problems. Methods/Statistical Analysis: This work projected a novel Fuzzy Classification Algorithm to train Fuzzy Inference System to classify and predict soil profile data with soil Total Porosity and proposed a fuzzy-genetic hybrid based clustering algorithm namely GENetic algorithm for SOIL (GENSOIL) profile data. Soil samples were collected from International Soil Reference and Information Centre. Findings: Fuzzy Rule Base has been developed and using in a proposed Fuzzy Classification Algorithm to train Fuzzy Inference System for soil profile data classification and prediction. The proposed algorithm classified the soil samples based on Fuzzy Rule Base using with Fuzzy Membership Function and hybrid with a randomized and optimized Genetic algorithm. A Novel Fuzzy-Genetic Hybrid Classification Algorithm compared with k-Means and Fuzzy C-Means are assessed on the basis of the time complexity of clustering. Applications/Improvements: The advantage of novel Fuzzy-Genetic Hybrid Classification Algorithm is their applicability in different types of related optimization problems with a superior speed of calculation and found solution very close to the best one.

Keywords

Classification, Fuzzy C-Means, Fuzzy Logic, Fuzzy Rule Base, Genetic, K-Means, Soil Profile

Full Text