Machine learning for hackers reading notes (12) model comparison

Library (' Ggplot2 ')
DF <-read.csv (' g:\\dataguru\\ml_for_hackers\\ml_for_hackers-master\\12-model_comparison\\data\\df.csv ')

#用glm

Logit.fit <-GLM (Label ~ X + y,family = binomial (link = ' logit '), data = DF)

Logit.predictions <-IfElse (Predict (Logit.fit) > 0, 1, 0)

Mean (with (df, logit.predictions = = Label))
#正确率 0.5156, with the same results as the guess.

Library (' e1071 ')

Svm.fit <-SVM (Label ~ X + Y, data = DF)

Svm.predictions <-IfElse (Predict (Svm.fit) > 0, 1, 0)

Mean (with (df, svm.predictions = = Label))

#改用SVM, correct rate 72%

Library ("reshape")

#df中的字段, X,Y,LABEL,LOGIT,SVM

DF <-Cbind (df,data.frame (Logit = IfElse (Predict (Logit.fit) > 0, 1, 0), SVM = IfElse (Predict (Svm.fit) > 0, 1, 0)))

#melt的结果, increase the field variable, where the value is LABEL,LOGIT,SVM, increment the field value, and take the corresponding value according to variable

#melt函数: Specify the variable, and the other remaining fields as a column, listing the corresponding values. Melt and cast, as if the opposite function

Predictions <-Melt (df, id.vars = C (' X ', ' Y '))

Ggplot (Predictions, AES (x = x, y = y, color = factor (value))) + geom_point () + facet_grid (variable ~.)

#如, the label is the real result, LGM completely useless, and SVM is recognized, but the edge is not

#SVM函数有个kernel参数, there are 4 values: Linear,polynomial,radial and sigmoid,4.

DF <-df[, C (' X ', ' Y ', ' Label ')]

Linear.svm.fit <-SVM (Label ~ X + Y, data = df, kernel = ' linear ')

With (DF, mean (Label = = IfElse (Predict (Linear.svm.fit) > 0, 1, 0)))

Polynomial.svm.fit <-SVM (Label ~ X + Y, data = df, kernel = ' polynomial ')

With (DF, mean (Label = = IfElse (Predict (Polynomial.svm.fit) > 0, 1, 0)))

Radial.svm.fit <-SVM (Label ~ X + Y, data = df, kernel = ' radial ')

With (DF, mean (Label = = IfElse (Predict (Radial.svm.fit) > 0, 1, 0)))

Sigmoid.svm.fit <-SVM (Label ~ X + Y, data = df, kernel = ' sigmoid ')

With (DF, mean (Label = = IfElse (Predict (Sigmoid.svm.fit) > 0, 1, 0)))

DF <-Cbind (DF,

Data.frame (LINEARSVM = IfElse (Predict (Linear.svm.fit) > 0, 1, 0),

POLYNOMIALSVM = IfElse (Predict (Polynomial.svm.fit) > 0, 1, 0),

RADIALSVM = IfElse (Predict (Radial.svm.fit) > 0, 1, 0),

SIGMOIDSVM = IfElse (Predict (Sigmoid.svm.fit) > 0, 1, 0)))

Predictions <-Melt (df, id.vars = C (' X ', ' Y '))

Ggplot (Predictions, AES (x = x, y = y, color = factor (value))) + geom_point () + facet_grid (variable ~.)

#如, linear and polynomial are useless, radial is OK, sigmoid is strange

#svm有个参数叫degree, look at the effect.

Polynomial.degree3.svm.fit <-SVM (Label ~ X + Y, data = df, kernel = ' polynomial ', degree = 3)

With (DF, mean (Label! = IfElse (Predict (Polynomial.degree3.svm.fit) > 0, 1, 0)))

Polynomial.degree5.svm.fit <-SVM (Label ~ X + y,data = Df,kernel = ' polynomial ', degree = 5)

With (DF, mean (Label! = IfElse (Predict (Polynomial.degree5.svm.fit) > 0, 1, 0)))

Polynomial.degree10.svm.fit <-SVM (Label ~ X + y,data = Df,kernel = ' polynomial ', degree = 10)

With (DF, mean (Label! = IfElse (Predict (Polynomial.degree10.svm.fit) > 0, 1, 0)))

Polynomial.degree12.svm.fit <-SVM (Label ~ X + Y, data = df, kernel = ' polynomial ', degree = 12)

With (DF, mean (Label! = IfElse (Predict (Polynomial.degree12.svm.fit) > 0, 1, 0)))

DF <-df[, C (' X ', ' Y ', ' Label ')]

DF <-Cbind (DF,

Data.frame (DEGREE3SVM = IfElse (Predict (Polynomial.degree3.svm.fit) > 0, 1, 0),

DEGREE5SVM = IfElse (Predict (Polynomial.degree5.svm.fit) > 0, 1, 0),

DEGREE10SVM = IfElse (Predict (Polynomial.degree10.svm.fit) > 0, 1, 0),

DEGREE12SVM = IfElse (Predict (Polynomial.degree12.svm.fit) > 0, 1, 0)))

Predictions <-Melt (df, id.vars = C (' X ', ' Y '))

Ggplot (Predictions, AES (x = x, y = y, color = factor (value))) + geom_point () + facet_grid (variable ~.)

#从图上看, the degreee is improved and the accuracy is increased, and there is an overfitting problem, so when using the polynomial kernel function, the degree is cross-validated

#接下来研究一下SVM的cost参数

Radial.cost1.svm.fit <-SVM (Label ~ X + Y, data = df, kernel = ' radial ', cost = 1)

With (DF, mean (Label = = IfElse (Predict (Radial.cost1.svm.fit) > 0, 1, 0)))

Radial.cost2.svm.fit <-SVM (Label ~ X + Y, data = df, kernel = ' radial ', cost = 2)

With (DF, mean (Label = = IfElse (Predict (Radial.cost2.svm.fit) > 0, 1, 0)))

Radial.cost3.svm.fit <-SVM (Label ~ X + Y, data = df, kernel = ' radial ', cost = 3)

With (DF, mean (Label = = IfElse (Predict (Radial.cost3.svm.fit) > 0, 1, 0)))

Radial.cost4.svm.fit <-SVM (Label ~ X + Y, data = df, kernel = ' radial ', cost = 4)

With (DF, mean (Label = = IfElse (Predict (Radial.cost4.svm.fit) > 0, 1, 0)))

DF <-df[, C (' X ', ' Y ', ' Label ')]

DF <-Cbind (DF,

Data.frame (COST1SVM = IfElse (Predict (Radial.cost1.svm.fit) > 0, 1, 0),

COST2SVM = IfElse (Predict (Radial.cost2.svm.fit) > 0, 1, 0),

COST3SVM = IfElse (Predict (Radial.cost3.svm.fit) > 0, 1, 0),

COST4SVM = IfElse (Predict (Radial.cost4.svm.fit) > 0, 1, 0)))

Predictions <-Melt (df, id.vars = C (' X ', ' Y '))

Ggplot (Predictions, AES (x = x, y = y, color = factor (value))) + geom_point () + facet_grid (variable ~.)

#,cost parameter value promotion makes the effect more and more poor, the change is very small, only through the edge data to perceive that the effect is getting worse

#再来看SVM的参数gamma

Sigmoid.gamma1.svm.fit <-SVM (Label ~ X + Y, data = df, kernel = ' sigmoid ', gamma = 1)

With (DF, mean (Label = = IfElse (Predict (Sigmoid.gamma1.svm.fit) > 0, 1, 0)))

Sigmoid.gamma2.svm.fit <-SVM (Label ~ X + Y, data = df, kernel = ' sigmoid ', gamma = 2)

With (DF, mean (Label = = IfElse (Predict (Sigmoid.gamma2.svm.fit) > 0, 1, 0)))

Sigmoid.gamma3.svm.fit <-SVM (Label ~ X + Y, data = df, kernel = ' sigmoid ', gamma = 3)

With (DF, mean (Label = = IfElse (Predict (Sigmoid.gamma3.svm.fit) > 0, 1, 0)))

Sigmoid.gamma4.svm.fit <-SVM (Label ~ X + Y, data = df, kernel = ' sigmoid ', gamma = 4)

With (DF, mean (Label = = IfElse (Predict (Sigmoid.gamma4.svm.fit) > 0, 1, 0)))

DF <-df[, C (' X ', ' Y ', ' Label ')]

DF <-Cbind (DF,

Data.frame (GAMMA1SVM = IfElse (Predict (Sigmoid.gamma1.svm.fit) > 0, 1, 0),

GAMMA2SVM = IfElse (Predict (Sigmoid.gamma2.svm.fit) > 0, 1, 0),

GAMMA3SVM = IfElse (Predict (Sigmoid.gamma3.svm.fit) > 0, 1, 0),

GAMMA4SVM = IfElse (Predict (Sigmoid.gamma4.svm.fit) > 0, 1, 0)))

Predictions <-Melt (df, id.vars = C (' X ', ' Y '))

Ggplot (Predictions, AES (x = x, y = y, color = factor (value))) + geom_point () + facet_grid (variable ~.)

#变弯曲了

#SVM介绍完毕, it means to meet the data set to adjust the parameters, the following comparison of the performance of SVM,GLM and KNN

Load (' G:\\dataguru\\ml_for_hackers\\ml_for_hackers-master\\12-model_comparison\\data\\dtm. RData ')

Set.seed (1)

#一半训练, half test

Training.indices <-Sort (sample (1:nrow (DTM), round (0.5 * nrow (DTM))))

Test.indices <-which (1:nrow (DTM)%in% training.indices)

Train.x <-dtm[training.indices, 3:ncol (DTM)]

TRAIN.Y <-dtm[training.indices, 1]

Test.x <-dtm[test.indices, 3:ncol (DTM)]

Test.y <-dtm[test.indices, 1]

RM (DTM)

Library (' Glmnet ')

Regularized.logit.fit <-glmnet (train.x, train.y, family = C (' binomial '))

lambdas <-Regularized.logit.fit$lambda

Performance <-Data.frame ()

For (lambda in lambdas)

{

Predictions <-predict (Regularized.logit.fit, test.x, s = lambda)

Predictions <-as.numeric (Predictions > 0)

MSE <-mean (predictions! = TEST.Y)

Performance <-Rbind (performance, data.frame (lambda = lambda, MSE = MSE))

}

Ggplot (Performance, AES (x = Lambda, y = MSE)) + geom_point () + SCALE_X_LOG10 ()

#有两个lambda对应的错误率是最小的, we chose the bigger one, because it means stronger regularization.

Best.lambda <-with (performance, Max (Lambda[which (MSE = = min (MSE))))

#算一下mse, 0.068

MSE <-with (subset (performance, Lambda = = Best.lambda), MSE)

#下面试一下SVM

Library (' e1071 ')

#这一步时间很长, because the data set is large, the linear kernel function takes a long time

Linear.svm.fit <-SVM (train.x, train.y, kernel = ' linear ')

Predictions <-predict (Linear.svm.fit, test.x)

Predictions <-as.numeric (Predictions > 0)

MSE <-mean (predictions! = TEST.Y)

Mse

#0.128, Error rate 12%, higher than GLM. To achieve optimal results, you should try different cost super-parameters

Radial.svm.fit <-SVM (train.x, train.y, kernel = ' radial ')

Predictions <-predict (Radial.svm.fit, test.x)

Predictions <-as.numeric (Predictions > 0)

MSE <-mean (predictions! = TEST.Y)

Mse

#错误率, 0.1421538, is higher than just now, so it is known that the radial kernel function is not effective, so the boundary may be linear. So the GLM effect will be better.

#下面试一下KNN, KNN is good for nonlinear effect

Library (' class ')

Knn.fit <-KNN (train.x, Test.x, train.y, k = 50)

Predictions <-as.numeric (As.character (Knn.fit))

MSE <-mean (predictions! = TEST.Y)

Mse

#错误率0.1396923, it is true that it is possible to have a linear model, and try to see which K works best.

Performance <-data.frame ()

for (k in SEQ (5, a by = 5))

{

Knn.fit <-KNN (train.x, Test.x, train.y, k = k)

Predictions <-as.numeric (As.character (Knn.fit))

MSE <-mean (predictions! = TEST.Y)

Performance <-Rbind (performance, data.frame (k = k, MSE = MSE))

}

BEST.K <-with (performance, k[which (MSE = = min (MSE)))

Best.mse <-with (subset (performance, K = = BEST.K), MSE)

Best.mse

#错误率降到0.09169231,KNN effect between GLM and SVM

#因此, the best choice is GLM.

Machine learning for hackers reading notes (12) model comparison