Improving a Flower Pollination Algorithm to Demonstrate Data Upgrading to Deep Data

Abstract

Most computational tasks if not all depends highly on the data formula, as well as the techniques used to achieve these tasks like clustering, classification and regression. Therefor most recent research focus on using data preprocessing as an initial stage. In this research a framework is designed and built that depends on two approaches, the first one is transforming the data into a new formula called deep data through seven structural steps instead of data preprocessing in order to get a data formula that is more robust, reliable and general form. The second approach resides in using the improved flower pollination algorithm (IFPA) relying on local and global search that guaranties the dual behavior to achieve exploitation and exploration behaviors which make the algorithm more flexible in dealing with different tasks like clustering ,classification and regression. The proposed framework is based on obtaining a data formula that is deep data as new solution for processing unstructured and unbalance data in different types textual, numerical and descriptive data types which gave the algorithm an explicit flexibility to manage various tasks. The results shows obvious superiority of the proposed framework in comparison to traditional methods without using Deep data, improved FPA or both, also the proposed framework proved robustness in dealing with clustering, classification and regression as the experimental results showed