| Data mining technology is an interdisciplinary, an interdisciplinary, involving a number of areas such as databases, statistics, artificial intelligence and machine learning and so on. Data mining, also known as knowledge discovery in database, it get effective, potentially useful and ultimately understandable patterns of non-trivial process from a mass of data. Obviously, data mining is extracted or "mining" knowledge from the large amount of data. Classification is a very important research subject of data mining technology, a model or function of the same class can be extracted from the data set by Classification, And the data set for each object of unknown class attributed to a known object class. Recently classification algorithms are mainly statistical classification and neural networks, decision trees, etc. Different classification algorithm will produce different classifiers, the classifier is good or bad directly affect the efficiency and accuracy of data mining. Therefore, when the huge mass of data is classified, it is very important to choose the most effective classification algorithm.However,the effect of classification generally relate to the characteristics of the datas. some datas have large noise,some datas have a missing value,some datas have a sparse distribution,some data fields or attributes have strong correlation, some data’s attributes are discrete,but another data attributes are continuous or hybrid. There isn’t a method which is suitable for all different data. The major task of this paper including summarizing the research status of the ID3algorithm, Naive Bayes algorithm, BP neural network algorithm, base on deeply understanding of three kinds of algorithm, do comparation with four data sets on the aspect of forecast accuracy, the time to establish classification model, finally summarize the advantages and disadvantage of the three algorithms., and try to look forward to the future of those algorithms. |
PDF Full Text Request