在 R 中使用 lda 进行预测：警告消息：'newdata'有 1600 行，但找到的变量有 200 行

Question

我是R的新手。我正在尝试使用LDA在生成的网格中对所有点进行分类。训练集是使用rmvnorm(n,mean,sigma)随机生成的两个点组。这是我的代码：`

 # number of samples
n=100;
# parameters: G2
meanG1 = matrix( 
  c(2, 2), # the data elements 
  nrow=1,              # number of rows 
  ncol=2,              # number of columns 
  byrow = TRUE)        # fill matrix by rows 
sigmaG1 = matrix( 
  c(1,0,0,1), # the data elements 
  nrow=2,              # number of rows 
  ncol=2,              # number of columns 
  byrow = TRUE)        # fill matrix by rows 
library(mvtnorm)  
# Generating a matrix G1 with norm distribution 
G1 = rmvnorm(n, meanG1, sigmaG1)
G1[,3]=1
# parameters: G2
meanG2 = matrix( 
  c(0, 0), # the data elements 
  nrow=1,              # number of rows 
  ncol=2,              # number of columns 
  byrow = TRUE)        # fill matrix by rows 
sigmaG2 = matrix( 
  c(1,0.75,0.75,1), # the data elements 
  nrow=2,              # number of rows 
  ncol=2,              # number of columns 
  byrow = TRUE)        # fill matrix by rows 
# # Generating a matrix G2 with norm distribution
G2 = rmvnorm(n, meanG2, sigmaG2)
# adding a column as a label = 1 to G1 matrix 
G1 = cbind(G1, 1 )
# adding a column as a label = 2 to G2 matrix 
G2 = cbind(G2, 2 )    
# Concatenate both matrices
G = rbind(G1,G2)    
# Transforming Matrix into dataFrame
bothGroupsWithLabel <- as.data.frame(G)
# Shuffling data row-wise
bothGroupsWithLabel <- bothGroupsWithLabel[sample(nrow(bothGroupsWithLabel)),]
# plotting the generated matrices
plot(c(G1[,1]),c(G1[,2]),col="red")
points(c(G2[,1]),c(G2[,2]),col="blue")
# Generating a grid
K = 40;
seqx1 = seq(min(G1[,1]),max(G1[,1]),length = K)
seqx2 = seq(min(G1[,2]),max(G1[,2]),length = K)
myGrid = expand.grid(z1=seqx1,z2=seqx2);
plot(myGrid[,1],myGrid[,2])
library(MASS)  
# Creating a model
model.lda = lda(bothGroupsWithLabel[,3] ~bothGroupsWithLabel[,1]+bothGroupsWithLabel[,2] , data = bothGroupsWithLabel);
Ypred = predict(model.lda, newdata=myGrid);
Ypredgrid = Ypred$class

这是我数据bothGroupsWithLabel V1 V2 V3 69 2.0683949 0.5779272 1 53 2.1261046 2.0420350 1 118 -1.4502033 -1.4775360 2 148 1.1705251 1.5437296 2 195 0.3100763 -0.2594026 2 40 1.8573633 3.7717020 1

的一部分

和 myGrid z1 z2 1 0.1048024 -0.2034172 2 0.2227540 -0.2034172 3 0.3407055 -0.2034172 4 0.4586571 -0.2034172 5 0.5766086 -0.2034172 6 0.6945602 -0.2034172

我的网格由40*40点组成，因此myGird数据框的大小为1600行和2列。数据框架bothGroupsWithLabel由200行和3列组成，前两列是点的坐标，第三列用于标签。我的问题是当我致电predict(model.lda, newdata=myGrid)时，我会收到此警告消息： Warning message: 'newdata' had 1600 rows but variables found have 200 rows我在这里想念什么？谁能帮帮我吗？

Answer 1

问题是您生成模型的方式。使用公式和data=...时，最好只使用变量名。为了使其工作，您还必须使newdata中的变量名称匹配。因此，当您创建myGrid添加行时：

names(myGrid) = c("V1", "V2")

然后使您的最后几行是：

model.lda = lda(V3 ~ V1 + V2 , data = bothGroupsWithLabel);
Ypred = predict(model.lda, newdata=myGrid);
Ypredgrid = Ypred$class

应该得到您想要的东西。