人工智能教学资料 Classification by Weka.pdfVIP

Artificial intelligence report Group Number: 38 1.Specification Our group choose to complete the project which is implementing a classification model based on the decision tree. The method that we used is Weka. However, because we have not learned much about the AI, what we can do is just analyzing the different decision tree algorithm with different data set. 2.Experiment 2.1 Data Set We choose three data sets to analyze the decision tree algorithm and the information about these data sets are show on the table below: Table 1. Data set Data set Instances Attributes weather.normal 14 5 Glass 214 10 german-credit 1000 21 As we can see that, the instances and attributes of each data set are both increasing. The reason why we choose these data sets is that we want to find the performance of different decison tree based on various data sets. 2.2 Decision tree algorithm In the project, we choose two classical algorithm that the teacher introduced in the class, that is the J48(C4.5) and SimpleCar(CART).Here we cannot use the ID3 algorithm, because the data set we used is continuous, the ID3 which can only used for discrete data. Now let’s take a simple look to these algorithm. 2.2.1 J48(C4.5) J48 is an improvement of ID3,which used the information gain radio to construct the decision tree, but not the information gain used in ID3. Suppose the data set is S and A is the attribute. The split information is: The information gain ratio is: The attribute which provides the largest information gain ratio is chosen to split the node. The J48(C4.5) can not only used for discrete data, but also the continuous data. 2.2.2 SimpleCar(CART) CART algorithm split each sub data set into two parts, so the CART decision tree is a binary tree. The CART algorithm used gini index to construct the decision tree. If a data set D contains examples from n classes, gini index, gini(D) is defined as: If a data set D is split