Optimization Technique For Feature Selection And Classification Using Support Vector Machine

In this module, the class label for the testing data is predicted. The n – dimensional feature vector for the testing data is converted from query tree of testing data in the manner similar to the data pre – processing phase. The SQLIA classifier determines the new testing feature vector is normal or malicious, by using optimized SVM classification model.

The major issue filter based feature subset methods are producing feature subset with maximum relevancy and minimum redundancy. The subset of features ‗S‘ are identified based on selecting feature subset at the multiple level . Finally, a subset of 19 % features are produced from the gives the better classifier performance when compared with other feature subsets methods . Extensive experiments are conducted on image data sets and shows that our framework for selecting optimal feature subset produced better efficiency and accuracy. The future scope of the work is to annotate the image regions and can retrieve relevant images based on image

We have used support vector machine (SVM) for classification task. We have used RBF kernel for training the classifier. 10 fold cross-validation is used for determining cost parameter C and best kernel width for RBF kernel function. If we perform classification without any feature selection or feature extraction then the accuracy is 48.99% and 65.82% for AVIRIS and HYDICE image respectively which is very poor and it highly motivates us to apply feature reduction technique. In table II we have shown the classification accuracy for each of the pair of class for PCA, MI and PCA-QMI.

Initially [4] the images similar to the query image are extracted from a large group of medical images. Then the search is by accelerating the retrieval process with the help of Support Vector Machine (SVM) classifier. The performance of the retrieval system is enhanced by adapting the subjective feedback method of Support Vector Machine (SVM) classifier. The performance of the retrieval system is enhanced by adapting the subjective feedback method.

Ensemble techniques have been more popular than single model [1]. In this technique more than one classifier is used for classification with higher efficiency. Each classifier in the classification model is trained on different data chunks. With the help of advanced data streaming technologies [2], we are now able to collect large volume of data for different application domains. For example credit card transaction, network traffic monitoring etc. the presence of irrelevant and redundant data slows down the learning algorithms [3] [4]. By removing or ignoring irrelevant and redundant feature, prediction performance and computational efficiency can be improved. Multiclass miner works with dynamic feature vector and detects novel classes. It is a combination of OLINDDA and FAE approach. OLINDDA and FAE are used to detect novel classes and to classify data chunks

Data Mining is the non-trivial extraction of potentially useful information about data. In other words, Data Mining extracts the knowledge or interesting information from large set of structured data that are from different sources. There are various research domains in data mining specifically text mining, web mining, image mining, sequence mining, process mining, graph mining, etc. Data mining applications are used in a range of areas such as it is used for financial data analysis, retail and telecommunication industries, banking, health care and medicine. In health care, the data mining is mainly used for disease prediction. In data mining, there are several techniques have been developed and used for predicting the diseases that includes data preprocessing, classification, clustering, association rules and sequential patterns. This paper analyses the performance of two classification techniques such as Bayesian and Lazy classifiers for hepatitis dataset. In Bayesian classifier there are two algorithms namely BayesNet and NaiveBayes. In Lazy classifier we have two algorithms namely IBK and KStar. Comparative analysis is done by using the WEKA tool.It is open source software which consists of the collection of machine learning algorithms for data mining tasks.

The use of the support vector machine (SVM) as a classifier for HSI applications has been shown to be robust and highly accurate [5]–[7]. The samples or pixels are evaluated in SVM by means of their respective features or spectral bands, which can contribute to more robust discrimination as they include information from different spectral wavelengths. However, HSI data are usually prone to noise, which can reduce the discrimination ability limiting the accuracy in classification tasks. For that reason, there is great interest for a potential decomposition of the spectral profiles into components in such a way that noise could be removed or mitigated by avoiding particular components with high noisy content. In this decomposition context, a particularly interesting research area is the use of the empirical mode decomposition (EMD) technique applied in 1-D to the spectral profile of the pixels as briefly evaluated in [8]. The EMD is the main part of the Hilbert–Huang transform, an algorithm for the analysis of nonlinear and nonstationary time series [9], [10]. EMD decomposes a 1-D signal into a few components called intrinsic mode functions (IMFs) and has been widely used in processing signal applications such as speech recognition [11]. The reconstruction of the 1-D signal by some of their IMFs would provide an alternative signal that could be easier to classify. Although this was the objective, an evaluation in [8] showed no improvement at all but

When unknown word percentage (Shown in Table 2) is taken in to consideration, SVM based tagger has provided the highest accuracies when there are more unknown words (E2 and E4), whereas CRF has provided the highest accuracies when there is a less number of unknown words in testing data (E1, E3 and E5). This shows that SVM is more robust to unknown words.Moreover, when testing is done using a different domain (E4 and E5), SVM and CRF based taggers have provided the highest accuracies in E4 and E5, respectively. This confirms that SVM and CRF based taggers are more robust to domain adaptation.

We have used the latest version of LibSVM-3.20 [39] library for our SVM work. The Radial Basis Function (RBF) Gaussian Kernel is utilized as an SVM kernel function. The similarity-reduced IEDB dataset under test consists of twelve alleles of binders and non-binders peptides. The binders peptides represent our positive instances while the non-binder peptides represent the negative instances. We have selected balanced x-instances from each allele (i.e., x-binders and x-non-binders), where x is the minimum count number of all binders and non-binders peptides in each allele as shown in Table 1. The HLA-DRB1*0101 allele from Wang benchmark dataset also treated and prepared like IEDB dataset. We run our

This classifier is derived from Naïve Bayes conditional probability. This is suitable for datasets having less number of attributes.

Correlation Coefficient approach evaluates how well an individual feature contributes to the separation of classes. Ranking criteria is used to rank all features using their mean and standard deviation for all the samples of both classes. The correlation coefficient is successfully reduced the number of features and also kept good classification accuracy (Zong-Xia et.al,2006).

Weka (Waikato Environment for Knowledge Analysis) is a collection of machine learning algorithms developed on Java platform by University of Waikato, New Zealand. It is a data analyzing tool used for data modeling and predictive analysis. It is licensed under GNU General Public License. As it is implemented in Java it is platform independent. It is a comprehensive collection of various data preprocessing and modeling techniques. Some of the operations performed using Weka like classification, clustering etc. are discussed here. It provides implementation of various algorithms which can be applied on any kind of datasets.

In some cases, redundant features can lead to noisy data that distract the learning algorithm and degrade the accuracy of the IDS through which, training and testing processes will be slowed down. Significant features are confessed to have a high significance on the performance of the classifiers. And handling appropriate feature selection methods renders the models to make them feasible to construe, reducing the training times and augment the generalization [10] [11]. Filtering approach is used as a robust one in building IDS, a set of features is chosen which are treated as most effective correlating to the classification procedure [12].

Support Vector Machine (SVM) is primarily a classier method that performs classification tasks by constructing hyperplanes in a multidimensional space that separates cases of different class labels. SVM supports both regression and classification tasks and can handle multiple continuous and categorical variables. For categorical variables a dummy variable is created with case values as either 0 or 1. Thus, a categorical dependent variable consisting of three levels, say (A, B, C), is represented by a set of three dummy variables:

Abstract—The field of information extraction and retrieval has grown exponentially in the last decade. Sentiment analysis is a task in which you identify the polarity of given text using text processing and classification. There are various approaches in the task of classification of text into various classes. Use of particular algorithms depends on the kind of input provided. Analyzing and understanding when to use which algorithm is an important aspect and can help in improving accuracy of results.