文本分类（Text categorization）.docVIP

文本分类（Text categorization） Commonly used classification algorithm for: Decision trees, Rocchio, naive Bayes, neural networks, support vector machines, linear least squares fitting, kNN, genetic algorithms, maximum entropy, Generalized, Instance, Set, etc.. Here, pick only a few of the most representative algorithms, Kan kan. Rocchio algorithm The Rocchio algorithm should be the first and most intuitive solution for people to think about text categorization problems. The basic idea is to take a class in the average value of a sample document (such as all the Sports category in the document word Basketball the number of occurrences of a mean value, then the referee takes the average value, in order to do so), you can get a new vector image known as the center of mass, became the category centroid vector is the most representative said. When the new document needs to be judged, compare how much the new document is similar to the centroid (eight points, judging the distance between them) to determine that the new document does not belong to this class. A slightly modified Rocchio algorithm does not consider belong to this category of documents (called positive samples), also consider the document data does not belong to this category (called negative samples), the calculated centroid is as close as possible to the sample and try to stay away from negative samples. The Rocchio algorithm makes two deadly assumptions, making its performance surprisingly poor. One is that it around a category of documents only gathered in a mass, the actual situation is not so (this data is called linear inseparable); two is it is assumed that the training data is absolutely correct, because it does not have any quantitative measure of whether the sample contains the mechanism of noise, so it is no resistance to the wrong data. But the Rocchio classifier is very intuitive, easy to be understood, algorithm is simple, there are still some value in use, is often used for the comparison of diffe