The Research On Fuzzy Theory Biclustering Of Gene Expression Data

With the development of biological gene sequencing, the gene expression data are more and more complex. So, biclustering become a new hotspot which be used in mining the data. Biclustering is rows and columns to clustering at the same time, then it can clustering in local unlike traditional clustering. So biclustering is more and more valued and worthy research in bioinformatics.In this paper, we use the method of fuzzy theory to improve Cheng and Church algorithm and make the result more accurate in large data. Firstly, we try to join comprehensive evaluation standard in add or remove rows and columns of the algorithm, it can be more accurate to selection. Secondly, using maximize fuzzy random variable characteristics and judgment matrix to make the original data will not lost in iterations. Finally, preprocessing the gene expression, and select a large data and a small data to compared with the result of the experiment. The method mainly joined comprehensive evaluation and so on, and make the data will not be lost in iterations. Besides, the experiment shows the improved algorithm is better in large data, which is both average square residues and average capacity.Base on the improved algorithm that we use multi-objective optimization to further improved, and multi-objective optimal considering two standards: minimize squared residual and maximize capacity of biclustering. Then, using this method on gene expression data experiment and comparing the results from the previous experiments, it shows average square residues and average capacity of biclustering have been optimized. Finally, comparing the main trend biclustering algorithm and it shows several data results are in second place.