Feature Selection Optimization using Hybrid Relief-f with Self-adaptive Differential Evolution

While the genome sequences of different crops have been published and annotated, relatively little is known about the transcriptional networks that regulate gene expression. There are different ways of finding significant motifs in genome sequences, evolutionary motif finding algorithms being one of them. The work presented in this paper uses Differential Evolution (DE) Algorithm to discover new motifs in rice genome sequences and aims to provide an intuitive wizard to analyse the micro-array analysis of expression patterns (motifs) for the queried genes. The used DE algorithm involves crossover operation with fine tuning. Further, to search for already known motifs parallel version of adaptive hash based pattern matching algorithm has been implemented. The study is conducted on rice genome sequences for twelve chromosomes each consisting of approximately 70000 genome sequences. In this paper, experimental results provide a comparison of serial and parallel version of the algorithm applied for the same dataset which shows that parallel implementation improves the performance by almost 25-30% to search the patterns in genome sequences and provide a list of newly discovered significant motifs.

Section: Methodsmentioning

confidence: 99%

“…The selection of the balance of exploration and exploitations. There are other variants of DE [25][26][27][28] available in literature for various problem domains, where the variant differs interns of mutant strategy, objective functions and decision variables.…”

Section: Literature Reviewmentioning

confidence: 99%

An Intuitive Wizard to Identify Pattern in Similarity Expressed Gene Set of Rice

Barrow¹,

Naryanan²

2018

“…Data sets generally have missing values and other deficiencies. Irrelevant features can be identified to reduce computational complexity [1][2][3][4]. Machine learning (ML) requires preprocessing to handle characteristic deficiencies such as missing and inconsistent values.…”

Section: Introductionmentioning

confidence: 99%

“…[5]. Other computational methods for dealing with the problem of missing values use the mechanism of evolution to find substitute values, such as genetic algorithm (GA), ant colony optimization (ACO), particle swarm optimization (PSO), artificial classification to optimize the parameters in a variety of applications [4,6,7]. Genetic algorithms can find a diverse set of solutions with search techniques based on the evolutionary principles of natural selection and genetics because of their ability to search various regions in the solution space.…”

Section: Introductionmentioning

confidence: 99%

Optimization of Feature Selection Using Genetic Algorithm in Naïve Bayes Classification for Incomplete Data

Khotimah¹,

Miswanto²,

Suprajitno³

2020

In the case of high dimensional data with missing values, the process of collecting data from various sources may be miss accidentally, which affected the quality of learning outcomes. a large number of machine learning methods can be applied to explore the search area for imputation and selection of features and parameters. ML classification needs preprocessing with self-organizing map imputation (SOMI) before the imputation of missing values is done to improve the accuracy of the model. This study introduces a new approach that combines naïve Bayes classification (NBC) and genetic algorithm (GA) optimization procedures to effectively explore the search space based on a sample of experimental points. GA is a classification model approach based on the selection of features that cause computational problems, such as reduced dimensions, uncertainty and imbalanced data sets with various classes. In the experiment, preprocessing the data using SOMI yielded error results that were up to 10% for various data sets with missing data compared to other methods. In the SOMI-GANB hybrid model, the experimental results show that the proposed method can significantly improve accuracy by up to 90% compared to other imputation methods and without feature selection. SOMI can be used for homogeneous, heterogeneous and mixed data sets. The results from the experiment clearly showed that the proposed method could significantly increase the yield compared to the other imputation methods and without feature selection. The combination of GA and naïve Bayes classification was chosen because they are simple, easy-to-understand methods that are very effective in finding optimal solutions from a set of possible solutions. Naïve Bayes imputation had higher accuracy compared to neural network imputation.

“…Some are statistical based, such as correlation methods, chi-square and relief-F [6][7][8]. Another method is based on mutual information (MI) by utilizing information gain and entropy [9][10][11]. Statistical methods identify the relevance and redundancy of a feature based on the linear relationships between features, while MI identifies the relevance and redundancy of a feature based on its non-linear relationships.…”

Section: Introductionmentioning

confidence: 99%

Correlation and Symmetrical Uncertainty-Based Feature Selection for Multivariate Time Series Classification

Saikhu¹,

Arifin²,

Fatichah³

2019

Data relating to everyday phenomena can be recorded in the form of time series. Recently, modeling of time-series has become a topic of interest in data mining. Conducting an analysis of multivariate time series effectively is an essential task for decision-making activities in fields such as meteorology, medicine, and finance. Features selection is a key problem in the analysis of multivariate time series. Rainfall prediction, biomedical classification, pattern recognition, sensor network analysis and so on all have different input features. The problem is that these features have interdependencies and time-delay relationships. Currently, research on the selection of input features of these data still depends on whether they are linear or non-linear. In this paper, we propose a new integration strategy between Pearson Correlation and Symmetrical Uncertainty for relevant feature selection based on linear and non-linear relationships for multivariate time-series classification. We evaluated the goodness of fit of feature subsets using merit value. The meteorological data set was used to test the proposed method. The result showed that the method was able to reduce the number of features by 77.9% features and increase their merit value 2.25 times compared to no input features selection.