High dimension and small sample size is an inherent problem of gene expression datasets which makes the analysis process more complex. The present study has developed a novel learning scheme that... Show More

High dimension and small sample size is an inherent problem of gene expression datasets which makes the analysis process more complex. The present study has developed a novel learning scheme that encapsulates a hybrid evolutionary fuzzy-rough feature selection model with an adaptive neural net ensemble. Fuzzy-rough method deals with uncertainty and impreciseness of real valued gene expression dataset and evolutionary search concept optimizes the subset selection process. The efficiency of the hybrid-FRGSNN model is evaluated by the proposed neural net ensemble learning algorithm. Again to prove the learning capability of ensemble algorithm, performance of the component classifiers pairing with FR, GSNN and FRGSNN are compared with proposed hybrid-FRGSNN based ensemble model. In addition to this, efficiency of neural net ensemble is compared with two classical and one advanced ensemble learning algorithms.

Circulating Tumor Cells (CTCs) are cells that have shed into the vasculate from the primary tumor and circulate into the blood stream. In this proposed work, the major genes causing the breast... Show More

Circulating Tumor Cells (CTCs) are cells that have shed into the vasculate from the primary tumor and circulate into the blood stream. In this proposed work, the major genes causing the breast cancer is identified by the principle of Association Rule. The trained set and training set is made to upload on the data store. By associating each row of a training set to all the rows of the trained data is done and the report is generated. The Baum welch process is called for the estimation of actual probabilities and emission probabilities by calculating its log likelihood factor which gives the high Priority gene values that are responsible for the cause of cancer. Based on this cell category is splitted into three clusters such as carcinoma level, metastasis level and Kaposi sarcoma. On each cluster it finds the highest priority value in it and classifies into high, low and medium values. On extraction of these higher gene values yields the major responsible genes causing breast cancer. Finally, the obtained results are validated through hierarchical clustering.

A novel biclustering approach is proposed in this paper, which can be used to cluster data (like web data, gene expression data) into local pattern using MapReduce framework. The proposed... Show More

A novel biclustering approach is proposed in this paper, which can be used to cluster data (like web data, gene expression data) into local pattern using MapReduce framework. The proposed biclustering approach extracts the highly coherent bicluster using a correlation measure called Average Correlation Value measure. Furthermore, MapReduce based genetic algorithm is firstly used to the biclustering of web data. This method can avoid local convergence in the optimization algorithms mostly. The MSWeb dataset and MSNBC web usage data set are used to test the performance of new MapReduce based Evolutionary biclustering algorithm. The experimental study is carried out for comparison of proposed algorithm with traditional genetic algorithm in biclustering. The results reveal that novel proposed approach preforms better than existing evolutionary biclustering approach.

One of the prominent issues in Genetic Algorithm (GA) is premature convergence on local optima. This restricts the enhanced optimal solution searching in the entire search space. Population size is... Show More

One of the prominent issues in Genetic Algorithm (GA) is premature convergence on local optima. This restricts the enhanced optimal solution searching in the entire search space. Population size is one of the influencing factors in Genetic Algorithm. Increasing the population size will improvise the randomized searching and maintains the diversity in the population. It also increases its computational complexity. Especially in GA Biclustering (GABiC), the search should be randomized to find more optimal patterns. In this paper, a novel approach for population setup in MapReduce framework is proposed. The maximal population is split into population sets, and these groups will proceed searching in parallel using MapReduce framework. This approach is attempted for biclustering the gene expression dataset in this paper. The performance of this proposed work seems promising on comparing its results with those obtained from previous hybridized optimization approaches. This approach will also handle data scalability issues and applicable to the big data biclustering problems.