第九讲 分类与预测宣讲培训.ppt

第八讲 分类与预测;Classification and Prediction;分类(Classification): 预测分类标号(离散值或名词性词) 建立一个模型，基于训练集的数据和分类属性的值（类标识）来分类，并在新数据中使用该模型。 预测(Prediction): 连续值函数上的建模, 例如，预测未知或缺失的值 典型应用 信用度 目标市场 医疗诊断;分类的两个步骤;Classification Process (1): Model Construction;Classification Process (2): Use the Model in Prediction;决策树算法;;Attribute Selection: Information Gain;Computing Information-Gain for Continuous-Value Attributes;Gain Ratio for Attribute Selection (C4.5);Gini 指标 (CART, IBM IntelligentMiner);Gini index (CART, IBM IntelligentMiner);Comparing Attribute Selection Measures;Overfitting and Tree Pruning;决策树归纳基本算法的加强;大型数据库中的分类;可伸缩的决策树归纳方法;Classification and Prediction;贝叶斯分类;贝叶斯定理: Basics;Towards Na?ve Bayesian Classifier;Derivation of Na?ve Bayes Classifier ;Na?ve Bayesian Classifier: Training Dataset;Na?ve Bayesian Classifier: An Example;Avoiding the 0-Probability Problem;Na?ve Bayesian Classifier: Comments;贝叶斯网络;An Example;训练贝叶斯信念网络;Classification and Prediction;Using IF-THEN Rules for Classification;age?;Rule Extraction from the Training Data;How to Learn-One-Rule?;;Classification and Prediction;分类: 预测分类的类标识 典型应用 {credit history, salary}- credit approval ( Yes/No) {Temp, Humidity} -- Rain (Yes/No);线性分类;Discriminative Classifiers;Classification by Backpropagation;Neural Network as a Classifier;神经元（感知器）;A Multi-Layer Feed-Forward Neural Network ;How A Multi-Layer Neural Network Works?;Defining a Network Topology;Backpropagation;;Classification and Prediction;SVM—Support Vector Machines;SVM—History and Applications;SVM—General Philosophy;SVM—Margins and Support Vectors;SVM—When Data Is Linearly Separable;SVM—Linearly Separable;SVM—Linearly Inseparable;SVM—Kernel functions;Classification and Prediction;Associative Classification;Typical Associative Classification Methods;Classification and Prediction;Lazy vs. Eager Learning;Lazy Learner: Instance-Based Methods;The k-Nearest Neighbor Algorithm;Case-Based Reasoning (CBR);Classification and Prediction;Genetic Algorithms (GA);Rough Set Approa