An Improved Invasive Weed Optimization Algorithm For Text Feature Selection

Now the relevant technology of the computer and network develops rapidly, people face a huge number of text information. Everyone can sit home, knowing what's going on in the world. However, with the increase of resource, it is becoming more and more difficult to find what you need from so much text information resources. The data mining technology was born, and text classification is a very important research hotspot and key technology in data mining. In text classification, text feature selection is one of the key technologies and core problems, and it has great effect to improve the speed and accuracy of information retrieval.In 2006 invasion weeds algorithm was put forward by A.R.Mehrabian and C.Lucas who were from the university of Tehran, Iran for the first time. It was in a paper named A novel numerical optimization algorithm inspired from weed colonization which was published in the journal of EcologicalInformatics. It was inspired by weeds growing process and proposed a method based on numerical optimization algorithm. It mimicked the basic process of grass diffusion, growth, reproduction and competitive survival. It could keep population diversity in the early and middle levels of evolution, thus the search in the solution space could be more comprehensive. It searched around the excellent individual to converge to global optimal solution gradually in the later levels of evolution.Current technology is nowhere close to make computers be able to think like people, to read and understand the central idea of the text, then sum up and select the right text feature, and the current mainstream methods are concluded in accordance with a measuring function of each entry in the text of the correlation function values, choosing a few from the front of the descending sequence. As a result, some words whose related function value is low, but with more useful information is ignored.In order to be more effectively, and to improve the accuracy of text feature selection, on the basis of text feature selection method based on standard invasion weeds algorithm, a method based on a kind of improved invasion weeds algorithm is proposed, namely to introduce the niche thought to classify the population reproduction competition, increase species diversity, improve the global search capability of the algorithm, and in the late algorithm using adaptive niche to improve the convergence precision, so as to improve the accuracy of text feature selection. The simulation experiment comparing with other methods is in order to further confirm the possibility of invasion weeds algorithm combined with the text feature selection, and to perfect the invasion of weeds algorithm based on text feature selection methods.