【基础】常用的机器学习&数据挖掘知识点Basis(基础):MSE(MeanSquareError均方误差),LMS(LeastMeanSquare最小均方),LSM(LeastSquareMethods最小二乘法),MLE(MaximumLikelihoodEstimation最大似然估计),QP(QuadraticProgramming二次规划),CP(ConditionalProbability条件概率),JP(JointProbability联合概率),MP(MarginalProbability边缘概率),BayesianFormula(贝叶斯公式),L1/L2Regularization(L1/L2正则,以及更多的,现在比较火的L2.5正则等),GD(GradientDescent梯度下降),SGD(StochasticGradientDescent随机梯度下降),Eigenvalue(特征值),Eigenvector(特征向量),QR-decomposition(QR分解),Quantile(分位数),Covariance(协方差矩阵)。CommonDistribution(常见分布):DiscreteDistribution(离散型分布):BernoulliDistribution/Binomial(贝努利分布/二项分布),NegativeBinomialDistribution(负二项分布),MultinomialDistribution(多项式分布),GeometricDistribution(几何分布),HypergeometricDistribution(超几何分布),PoissonDistribution(泊松分布)ContinuousDistribution(连续型分布):UniformDistribution(均匀分布),NormalDistribution/GuassianDistribution(正态分布/高斯分布),ExponentialDistribution(指数分布),LognormalDistribution(对数正态分布),GammaDistribution(Gamma分布),BetaDistribution(Beta分布),DirichletDistribution(狄利克雷分布),RayleighDistribution(瑞利分布),CauchyDistribution(柯西分布),WeibullDistribution(韦伯分布)ThreeSamplingDistribution(三大抽样分布):Chi-squareDistribution(卡方分布),t-distribution(t-distribution),F-distribution(F-分布)DataPre-processing(数据预处理):MissingValueImputation(缺失值填充),Discretization(离散化),Mapping(映射),Normalization(归一化/标准化)。Sampling(采样):SimpleRandomSampling(简单随机采样),OfflineSampling(离线等可能K采样),OnlineSampling(在线等可能K采样),Ratio-basedSampling(等比例随机采样),Acceptance-RejectionSampling(接受-拒绝采样),ImportanceSampling(重要性采样),MCMC(MarkovChainMonteCarlo马尔科夫蒙特卡罗采样算法:Metropolis-Hasting&Gibbs)。Clustering(聚类):K-Means,K-Mediods,二分K-Means,FK-Means,Canopy,Spectral-KMeans(谱聚类),GMM-EM(混合高斯模型-期望最大化算法解决),K-Pototypes,CLARANS(基于划分),BIRCH(基于层次),CURE(基于层次),DBSCAN(基于密度),CLIQUE(基于密度和基于网格)Classification&Regression(分类&回归):LR(LinearRegression线性回归),LR(LogisticRegression逻辑回归),SR(SoftmaxRegression多分类逻辑回归),GLM(GeneralizedLinearModel广义线性模型),RR(RidgeRegression岭回归/L2正则最小二乘回归),LASSO(LeastAbsoluteShrinkageandSelectionatorOperatorL1正则最小二乘回归),RF(随机森林),DT(DecisionTree决策树),GBDT(GradientBoostingDecisionTree梯度下降决策树),CART(ClassificationAndRegressionTree分类回归树),KNN(K-NearestNeighborK近邻),SVM(SupportVectorMachine),KF(KernelFunction核函数PolynomialKernelFunction多项式核函数、GuassianKernelFunction高斯核函数/RadialBasisFunctionRBF径向基函数、StringKernelFunction字符串核函数)、NB(NaiveBayes朴素贝叶斯),BN(BayesianNetwork/BayesianBeliefNetwork/BeliefNetwork贝叶斯网络/贝叶斯信度网络/信念网络),LDA(LinearDiscriminantAnalysis/FisherLinearDiscriminant线性判别分析/Fisher线性判别),EL(EnsembleLearning集成学习Boosting,Bagging,Stacking),AdaBoost(AdaptiveBoosting自适应增强),MEM(MaximumEntropyModel最大熵模型)EffectivenessEvaluation(分类效果评估):ConfusionMatrix(混淆矩阵),Precision(精确度),Recall(召回率),Accuracy(准确率),F-score(F得分),ROCCurve(ROC曲线),AUC(AUC面积),LiftCurve(Lift曲线),KSCurve(KS曲线)。PGM(ProbabilisticGraphicalModels概率图模型):BN(BayesianNetwork/BayesianBeliefNetwork/BeliefNetwork贝叶斯网络/贝叶斯信度网络/信念网络),MC(MarkovChain马尔科夫链),HMM(HiddenMarkovModel马尔科夫模型),MEMM(MaximumEntropyMarkovModel最大熵马尔科夫模型),CRF(ConditionalRandomField条件随机场),MRF(MarkovRandomField马尔科夫随机场)。NN(NeuralNetwork神经网络):ANN(ArtificialNeuralNetwork人工神经网络),BP(ErrorBackPropagation误差反向传播)DeepLearning(深度学习):Auto-encoder(自动编码器),SAE(StackedAuto-encoders堆叠自动编码器:SparseAuto-encoders稀疏自动编码器、DenoisingAuto-encoders去噪自动编码器、ContractiveAuto-encoders收缩自动编码器),RBM(RestrictedBoltzmannMachine受限玻尔兹曼机),DBN(DeepBeliefNetwork深度信念网络),CNN(ConvolutionalNeuralNetwork卷积神经网络),Word2Vec(词向量学习模型)。DimensionalityReduction(降维):LDALinearDiscriminantAnalysis/FisherLinearDiscriminant线性判别分析/Fisher线性判别,PCA(PrincipalComponentAnalysis主成分分析),ICA(IndependentComponentAnalysis独立成分分析),SVD(SingularValueDecomposition奇异值分解),FA(FactorAnalysis因子分析法)。TextMining(文本挖掘):VSM(VectorSpaceModel向量空间模型),Word2Vec(词向量学习模型),TF(TermFrequency词频),TF-IDF(TermFrequency-InverseDocumentFrequency词频-逆向文档频率),MI(MutualInformation互信息),ECE(ExpectedCrossEntropy期望交叉熵),QEMI(二次信息熵),IG(InformationGain信息增益),IGR(InformationGainRatio信息增益率),Gini(基尼系数),x2Statistic(x2统计量),TEW(TextEvidenceWeight文本证据权),OR(OddsRatio优势率),N-GramModel,LSA(LatentSemanticAnalysis潜在语义分析),PLSA(ProbabilisticLatentSemanticAnalysis基于概率的潜在语义分析),LDA(LatentDirichletAllocation潜在狄利克雷模型)AssociationMining(关联挖掘):Apriori,FP-growth(FrequencyPatternTreeGrowth频繁模式树生长算法),AprioriAll,Spade。RecommendationEngine(推荐引擎):DBR(Demographic-basedRecommendation基于人口统计学的推荐),CBR(Context-basedRecommendation基于内容的推荐),CF(CollaborativeFiltering协同过滤),UCF(User-basedCollaborativeFilteringRecommendation基于用户的协同过滤推荐),ICF(Item-basedCollaborativeFilteringRecommendation基于项目的协同过滤推荐)。SimilarityMeasure&DistanceMeasure(相似性与距离度量):EuclideanDistance(欧式距离),ManhattanDistance(曼哈顿距离),ChebyshevDistance(切比雪夫距离),MinkowskiDistance(闵可夫斯基距离),StandardizedEuclideanDistance(标准化欧氏距离),MahalanobisDistance(马氏距离),Cos(Cosine余弦),HammingDistance/EditDistance(汉明距离/编辑距离),JaccardDistance(杰卡德距离),CorrelationCoefficientDistance(相关系数距离),InformationEntropy(信息熵),KL(Kullback-LeiblerDivergenceKL散度/RelativeEntropy相对熵)。Optimization(最优化):Non-constrainedOptimization(无约束优化):CyclicVariableMethods(变量轮换法),PatternSearchMethods(模式搜索法),VariableSimplexMethods(可变单纯形法),GradientDescentMethods(梯