Teng Peng tengpeng

p < 0.05

tengpeng / a.rb

Created January 31, 2016 22:58

replace NA py

users.gender.replace('-unknown-', np.nan, inplace=True)

tengpeng / a.rb

Created January 31, 2016 22:59

drop NA py

	users_nan = (users.isnull().sum() / users.shape[0]) * 100
	users_nan[users_nan > 0].drop('country_destination')

tengpeng / a.rb

Created January 31, 2016 23:07

users[users.age > 122]['age'].describe()

tengpeng / a.rb

Created January 31, 2016 23:10

set na py

users.loc[users.age > 95, 'age'] = np.nan

tengpeng / a.rb

Created February 2, 2016 20:50

categorical to numerical

	## Convert categorical features into numerical features ##
	for (f in common_vars) {
	if (class(train[[f]]) == "character"){
	levels <- unique(c(train[[f]], test[[f]]))
	train[[f]] <- as.integer(factor(train[[f]], levels=levels))
	test[[f]] <- as.integer(factor(test[[f]], levels=levels))
	}
	}

tengpeng / a.rb

Created February 5, 2016 21:08

subset r

	nci = subset(nci, select = c("CNS", "RENAL", "BREAST", "NSCLC", "MELANOMA", "OVARIAN", "LEUKEMIA", "COLON"))
	nci = nci[, c("CNS", "RENAL", "BREAST", "NSCLC", "MELANOMA", "OVARIAN", "LEUKEMIA", "COLON")]
	nci = nci[, which(colnames(nci) %in% c("CNS", "RENAL", "BREAST", "NSCLC", "MELANOMA", "OVARIAN", "LEUKEMIA", "COLON"))]

tengpeng / a.rb

Created February 5, 2016 21:36

classify r

	result = if(any(result <= 1.5)) 1 else 2
	result = ifelse(result <= 1.5, 1, 2)

tengpeng / a.rb

Created February 5, 2016 23:40

binary classification r

prediction <- as.numeric(pred > 0.5)

tengpeng / a.rb

Created February 5, 2016 23:49

binary classification measure performance r

err <- mean(as.numeric(pred > 0.5) != test$label)

tengpeng / a.rb

Created February 6, 2016 01:32

binary classification xgboost r

	library(xgboost)

	X = as.matrix(train[,-58])
	y = train[,58]
	numberOfClasses <- max(y) + 1

	bst <- xgboost(data = X, label = y, max.depth = 2, eta = 1, nthread = 2, nround = 200, objective = "binary:logistic", verbose = 1)

	pred = predict(bst, data.matrix(test[,-58]))
	pred = as.numeric(pred > 0.5)