R語言 lightgbm 算法優(yōu)化：不平衡二分類問題（附代碼）
轉(zhuǎn)自 | 大數(shù)據(jù)文摘
作者｜蘇高生
本案例使用的數(shù)據(jù)為kaggle中“Santander Customer Satisfaction”比賽的數(shù)據(jù)。此案例為不平衡二分類問題，目標為最大化auc值（ROC曲線下方面積）。目前此比賽已經(jīng)結(jié)束。
競賽題目鏈接為：
https://www.kaggle.com/c/santander-customer-satisfaction?
2.建模思路
本文檔采用微軟開源的lightgbm算法進行分類，運行速度極快。具體步驟為:
讀取數(shù)據(jù)；
并行運算：由于lightgbm包可以通過設(shè)置相應參數(shù)進行并行運算，因此不再調(diào)用doParallel與foreach包進行并行運算；
特征選擇：使用mlr包提取了99%的chi.square；
調(diào)參：逐步調(diào)試lgb.cv函數(shù)的參數(shù)，并多次調(diào)試，直到滿意為止；
預測結(jié)果：用調(diào)試好的參數(shù)值構(gòu)建lightgbm模型，輸出預測結(jié)果；本案例所用程序輸出結(jié)果的ROC值為0.833386,已超過Private Leaderboard排名第一的結(jié)果(0.829072)。
3.lightgbm算法
由于lightgbm算法沒有給出具體的數(shù)學公式，因此此處不再介紹，如有需要，請查看github項目網(wǎng)址。
lightgbm算法具體介紹網(wǎng)址：
https://github.com/Microsoft/LightGBM
讀取數(shù)據(jù)
options(java.parameters = "-Xmx8g") ## 特征選擇時使用，但是需要在加載包之前設(shè)置
library(readr)
lgb_tr1 <- read_csv("C:/Users/Administrator/Documents/kaggle/scs_lgb/train.csv")
lgb_te1 <- read_csv("C:/Users/Administrator/Documents/kaggle/scs_lgb/test.csv")
數(shù)據(jù)探索
1.設(shè)置并行運算
library(dplyr)
library(mlr)
library(parallelMap)
parallelStartSocket(2)
2.數(shù)據(jù)各列初步探索
summarizeColumns(lgb_tr1) %>% View()
3.處理缺失值
impute missing values by mean and mode
imp_tr1 <- impute(
? ?as.data.frame(lgb_tr1),
? ?classes = list(
? ? ? ?integer = imputeMean(),
? ? ? ?numeric = imputeMean()
? ?)
)
imp_te1 <- impute(
? ?as.data.frame(lgb_te1),
? ?classes = list(
? ? ? ?integer = imputeMean(),
? ? ? ?numeric = imputeMean()
? ?)
)
處理缺失值后
summarizeColumns(imp_tr1$data) %>% View()
4.觀察訓練數(shù)據(jù)類別的比例–數(shù)據(jù)類別不平衡
table(lgb_tr1$TARGET)
5.剔除數(shù)據(jù)集中的常數(shù)列
lgb_tr2 <- removeConstantFeatures(imp_tr1$data)
lgb_te2 <- removeConstantFeatures(imp_te1$data)
6.保留訓練數(shù)據(jù)集與測試數(shù)據(jù)及相同的列
tr2_name <- data.frame(tr2_name = colnames(lgb_tr2))
te2_name <- data.frame(te2_name = colnames(lgb_te2))
tr2_name_inner <- tr2_name %>%
? ?inner_join(te2_name, by = c('tr2_name' = 'te2_name'))
TARGET = data.frame(TARGET = lgb_tr2$TARGET)
lgb_tr2 <- lgb_tr2[, c(tr2_name_inner$tr2_name[2:dim(tr2_name_inner)[1]])]
lgb_te2 <- lgb_te2[, c(tr2_name_inner$tr2_name[2:dim(tr2_name_inner)[1]])]
lgb_tr2 <- cbind(lgb_tr2, TARGET)
注：
1）由于本次使用lightgbm算法，故而不對數(shù)據(jù)進行標準化處理；
2）lightgbm算法運行效率極高，1GB內(nèi)不進行特征篩選也可以運行的極快，但是此處進行特征篩選，以進一步加快運行速率；
3）本案例直接進行特征篩選，未生成衍生變量，原因為：不知特征實際意義，不好隨機生成。
特征篩選–卡方檢驗
library(lightgbm)
1.試算最大權(quán)重值程序，后面將繼續(xù)優(yōu)化
grid_search <- expand.grid(
? ?weight = seq(1, 30, 2)
? ?## table(lgb_tr1$TARGET)[1] / table(lgb_tr1$TARGET)[2] = 24.27261
? ?## 故而設(shè)定weight在[1, 30]之間
)
lgb_rate_1 <- numeric(length = nrow(grid_search))
set.seed(0)
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr2$TARGET * i + 1) / sum(lgb_tr2$TARGET * i + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr2[, 1:300]),
? ? ? ?label = lgb_tr2$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc'
? ?)
? ?# 交叉驗證
? ?lgb_tr2_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?learning_rate = .1,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?lgb_rate_1[i] <- unlist(lgb_tr2_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr2_mod$record_evals$valid$auc$eval))]
}
library(ggplot2)
grid_search$perf <- lgb_rate_1
ggplot(grid_search,aes(x = weight, y = perf)) +
? ?geom_point()
從此圖可知auc值受權(quán)重影響不大,在weight=5時達到最大。
3.特征選擇
1)特征選擇
lgb_tr2$TARGET <- factor(lgb_tr2$TARGET)
lgb.task <- makeClassifTask(data = lgb_tr2, target = 'TARGET')
lgb.task.smote <- oversample(lgb.task, rate = 5)
fv_time <- system.time(
? ?fv <- generateFilterValuesData(
? ? ? ?lgb.task.smote,
? ? ? ?method = c('chi.squared')
? ? ? ?## 此處可以使用信息增益/卡方檢驗的方法，但是不建議使用隨機森林方法，效率極低
? ? ? ?## 如果有興趣，也可以嘗試IV值方法篩選
? ? ? ?## 特征工程決定目標值（此處為auc）的上限，可以把特征篩選方法作為超參數(shù)處理
? ?)
)
2)制圖查看
# plotFilterValues(fv)
plotFilterValuesGGVIS(fv)
3)提取99%的chi.squared（lightgbm算法效率極高，因此可以取更多的變量）
注：提取的X%的chi.squared中的X可以作為超參數(shù)處理。
fv_data2 <- fv$data %>%
? ?arrange(desc(chi.squared)) %>%
? ?mutate(chi_gain_cul = cumsum(chi.squared) / sum(chi.squared))
fv_data2_filter <- fv_data2 %>% filter(chi_gain_cul <= 0.99)
dim(fv_data2_filter) ## 減少了一半的自變量
fv_feature <- fv_data2_filter$name
lgb_tr3 <- lgb_tr2[, c(fv_feature, 'TARGET')]
lgb_te3 <- lgb_te2[, fv_feature]
4)寫出數(shù)據(jù)
write_csv(lgb_tr3, 'C:/users/Administrator/Documents/kaggle/scs_lgb/lgb_tr3_chi.csv')
write_csv(lgb_te3, 'C:/users/Administrator/Documents/kaggle/scs_lgb/lgb_te3_chi.csv')
算法
lgb_tr <- rxImport('C:/Users/Administrator/Documents/kaggle/scs_lgb/lgb_tr3_chi.csv')
lgb_te <- rxImport('C:/Users/Administrator/Documents/kaggle/scs_lgb/lgb_te3_chi.csv')
## 建議lgb_te數(shù)據(jù)在預測時再讀取，以節(jié)約內(nèi)存
library(lightgbm)
1.調(diào)試weight參數(shù)
grid_search <- expand.grid(
? ?weight = 1:30
)
perf_weight_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * i + 1) / sum(lgb_tr$TARGET * i + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc'
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?learning_rate = .1,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_weight_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
library(ggplot2)
grid_search$perf <- perf_weight_1
ggplot(grid_search,aes(x = weight, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
從此圖可知auc值在weight=4時達到最大,呈遞減趨勢。
2.調(diào)試learning_rate參數(shù)
grid_search <- expand.grid(
? ?learning_rate = 2 ^ (-(8:1))
)
perf_learning_rate_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 4 + 1) / sum(lgb_tr$TARGET * 4 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_learning_rate_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_learning_rate_1
ggplot(grid_search,aes(x = learning_rate, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
從此圖可知auc值在learning_rate=2^(-5) 時達到最大,但是 2^(-(6:3)) 區(qū)別極小，故取learning_rate = .125，提高運行速度。
3.調(diào)試num_leaves參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .125,
? ?num_leaves = seq(50, 800, 50)
)
perf_num_leaves_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 4 + 1) / sum(lgb_tr$TARGET * 4 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_num_leaves_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_num_leaves_1
ggplot(grid_search,aes(x = num_leaves, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
從此圖可知auc值在num_leaves=650時達到最大。
4.調(diào)試min_data_in_leaf參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .125,
? ?num_leaves = 650,
? ?min_data_in_leaf = 2 ^ (1:7)
)
perf_min_data_in_leaf_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 4 + 1) / sum(lgb_tr$TARGET * 4 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves'],
? ? ? ?min_data_in_leaf = grid_search[i, 'min_data_in_leaf']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_min_data_in_leaf_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_min_data_in_leaf_1
ggplot(grid_search,aes(x = min_data_in_leaf, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
從此圖可知auc值對min_data_in_leaf不敏感，因此不做調(diào)整。
5.調(diào)試max_bin參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .125,
? ?num_leaves = 650,
? ?max_bin = 2 ^ (5:10)
)
perf_max_bin_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 4 + 1) / sum(lgb_tr$TARGET * 4 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves'],
? ? ? ?max_bin = grid_search[i, 'max_bin']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_max_bin_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_max_bin_1
ggplot(grid_search,aes(x = max_bin, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
從此圖可知auc值在max_bin=2^10 時達到最大,需要再次微調(diào)max_bin值。
6.微調(diào)max_bin參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .125,
? ?num_leaves = 650,
? ?max_bin = 100 * (6:15)
)
perf_max_bin_2 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 4 + 1) / sum(lgb_tr$TARGET * 4 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves'],
? ? ? ?max_bin = grid_search[i, 'max_bin']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_max_bin_2[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_max_bin_2
ggplot(grid_search,aes(x = max_bin, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
從此圖可知auc值在max_bin=1000時達到最大。
7.調(diào)試min_data_in_bin參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .125,
? ?num_leaves = 650,
? ?max_bin=1000,
? ?min_data_in_bin = 2 ^ (1:9)
? ?
)
perf_min_data_in_bin_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 4 + 1) / sum(lgb_tr$TARGET * 4 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves'],
? ? ? ?max_bin = grid_search[i, 'max_bin'],
? ? ? ?min_data_in_bin = grid_search[i, 'min_data_in_bin']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_min_data_in_bin_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_min_data_in_bin_1
ggplot(grid_search,aes(x = min_data_in_bin, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
從此圖可知auc值在min_data_in_bin=8時達到最大,但是變化極其細微,因此不做調(diào)整。
8.調(diào)試feature_fraction參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .125,
? ?num_leaves = 650,
? ?max_bin=1000,
? ?min_data_in_bin = 8,
? ?feature_fraction = seq(.5, 1, .02)
? ?
)
perf_feature_fraction_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 4 + 1) / sum(lgb_tr$TARGET * 4 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves'],
? ? ? ?max_bin = grid_search[i, 'max_bin'],
? ? ? ?min_data_in_bin = grid_search[i, 'min_data_in_bin'],
? ? ? ?feature_fraction = grid_search[i, 'feature_fraction']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_feature_fraction_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_feature_fraction_1
ggplot(grid_search,aes(x = feature_fraction, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
從此圖可知auc值在feature_fraction=.62時達到最大,feature_fraction在[.60,.62]之間時，auc值保持穩(wěn)定,表現(xiàn)較好;從.64開始呈下降趨勢。
9.調(diào)試min_sum_hessian參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .125,
? ?num_leaves = 650,
? ?max_bin=1000,
? ?min_data_in_bin = 8,
? ?feature_fraction = .62,
? ?min_sum_hessian = seq(0, .02, .001)
)
perf_min_sum_hessian_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 4 + 1) / sum(lgb_tr$TARGET * 4 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves'],
? ? ? ?max_bin = grid_search[i, 'max_bin'],
? ? ? ?min_data_in_bin = grid_search[i, 'min_data_in_bin'],
? ? ? ?feature_fraction = grid_search[i, 'feature_fraction'],
? ? ? ?min_sum_hessian = grid_search[i, 'min_sum_hessian']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_min_sum_hessian_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_min_sum_hessian_1
ggplot(grid_search,aes(x = min_sum_hessian, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
從此圖可知auc值在min_sum_hessian=0.005時達到最大,建議min_sum_hessian取值在[0.002, 0.005]區(qū)間,0.005后呈遞減趨勢。
10.調(diào)試lamda參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .125,
? ?num_leaves = 650,
? ?max_bin=1000,
? ?min_data_in_bin = 8,
? ?feature_fraction = .62,
? ?min_sum_hessian = .005,
? ?lambda_l1 = seq(0, .01, .002),
? ?lambda_l2 = seq(0, .01, .002)
)
perf_lamda_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 4 + 1) / sum(lgb_tr$TARGET * 4 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves'],
? ? ? ?max_bin = grid_search[i, 'max_bin'],
? ? ? ?min_data_in_bin = grid_search[i, 'min_data_in_bin'],
? ? ? ?feature_fraction = grid_search[i, 'feature_fraction'],
? ? ? ?min_sum_hessian = grid_search[i, 'min_sum_hessian'],
? ? ? ?lambda_l1 = grid_search[i, 'lambda_l1'],
? ? ? ?lambda_l2 = grid_search[i, 'lambda_l2']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_lamda_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_lamda_1
ggplot(data = grid_search, aes(x = lambda_l1, y = perf)) +
? ?geom_point() +
? ?facet_wrap(~ lambda_l2, nrow = 5)
從此圖可知建議lambda_l1 = 0, lambda_l2 = 0
11.調(diào)試drop_rate參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .125,
? ?num_leaves = 650,
? ?max_bin=1000,
? ?min_data_in_bin = 8,
? ?feature_fraction = .62,
? ?min_sum_hessian = .005,
? ?lambda_l1 = 0,
? ?lambda_l2 = 0,
? ?drop_rate = seq(0, 1, .1)
)
perf_drop_rate_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 4 + 1) / sum(lgb_tr$TARGET * 4 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves'],
? ? ? ?max_bin = grid_search[i, 'max_bin'],
? ? ? ?min_data_in_bin = grid_search[i, 'min_data_in_bin'],
? ? ? ?feature_fraction = grid_search[i, 'feature_fraction'],
? ? ? ?min_sum_hessian = grid_search[i, 'min_sum_hessian'],
? ? ? ?lambda_l1 = grid_search[i, 'lambda_l1'],
? ? ? ?lambda_l2 = grid_search[i, 'lambda_l2'],
? ? ? ?drop_rate = grid_search[i, 'drop_rate']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_drop_rate_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_drop_rate_1
ggplot(data = grid_search, aes(x = drop_rate, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
從此圖可知auc值在drop_rate=0.2時達到最大,在0, .2, .5較好；在[0, 1]變化不大。
12.調(diào)試max_drop參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .125,
? ?num_leaves = 650,
? ?max_bin=1000,
? ?min_data_in_bin = 8,
? ?feature_fraction = .62,
? ?min_sum_hessian = .005,
? ?lambda_l1 = 0,
? ?lambda_l2 = 0,
? ?drop_rate = .2,
? ?max_drop = seq(1, 10, 2)
)
perf_max_drop_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 4 + 1) / sum(lgb_tr$TARGET * 4 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves'],
? ? ? ?max_bin = grid_search[i, 'max_bin'],
? ? ? ?min_data_in_bin = grid_search[i, 'min_data_in_bin'],
? ? ? ?feature_fraction = grid_search[i, 'feature_fraction'],
? ? ? ?min_sum_hessian = grid_search[i, 'min_sum_hessian'],
? ? ? ?lambda_l1 = grid_search[i, 'lambda_l1'],
? ? ? ?lambda_l2 = grid_search[i, 'lambda_l2'],
? ? ? ?drop_rate = grid_search[i, 'drop_rate'],
? ? ? ?max_drop = grid_search[i, 'max_drop']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_max_drop_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_max_drop_1
ggplot(data = grid_search, aes(x = max_drop, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
從此圖可知auc值在max_drop=5時達到最大,在[1, 10]區(qū)間變化較小。
二次調(diào)參
1.調(diào)試weight參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .125,
? ?num_leaves = 650,
? ?max_bin=1000,
? ?min_data_in_bin = 8,
? ?feature_fraction = .62,
? ?min_sum_hessian = .005,
? ?lambda_l1 = 0,
? ?lambda_l2 = 0,
? ?drop_rate = .2,
? ?max_drop = 5
)
perf_weight_2 <- numeric(length = nrow(grid_search))
for(i in 1:20){
? ?lgb_weight <- (lgb_tr$TARGET * i + 1) / sum(lgb_tr$TARGET * i + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[1, 'learning_rate'],
? ? ? ?num_leaves = grid_search[1, 'num_leaves'],
? ? ? ?max_bin = grid_search[1, 'max_bin'],
? ? ? ?min_data_in_bin = grid_search[1, 'min_data_in_bin'],
? ? ? ?feature_fraction = grid_search[1, 'feature_fraction'],
? ? ? ?min_sum_hessian = grid_search[1, 'min_sum_hessian'],
? ? ? ?lambda_l1 = grid_search[1, 'lambda_l1'],
? ? ? ?lambda_l2 = grid_search[1, 'lambda_l2'],
? ? ? ?drop_rate = grid_search[1, 'drop_rate'],
? ? ? ?max_drop = grid_search[1, 'max_drop']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?learning_rate = .1,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_weight_2[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
library(ggplot2)
ggplot(data.frame(num = 1:length(perf_weight_2), perf = perf_weight_2), aes(x = num, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
從此圖可知auc值在weight>=3時auc趨于穩(wěn)定, weight=7 the max
2.調(diào)試learning_rate參數(shù)
grid_search <- expand.grid(
? ?learning_rate = seq(.05, .5, .03),
? ?num_leaves = 650,
? ?max_bin=1000,
? ?min_data_in_bin = 8,
? ?feature_fraction = .62,
? ?min_sum_hessian = .005,
? ?lambda_l1 = 0,
? ?lambda_l2 = 0,
? ?drop_rate = .2,
? ?max_drop = 5
)
perf_learning_rate_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 7 + 1) / sum(lgb_tr$TARGET * 7 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves'],
? ? ? ?max_bin = grid_search[i, 'max_bin'],
? ? ? ?min_data_in_bin = grid_search[i, 'min_data_in_bin'],
? ? ? ?feature_fraction = grid_search[i, 'feature_fraction'],
? ? ? ?min_sum_hessian = grid_search[i, 'min_sum_hessian'],
? ? ? ?lambda_l1 = grid_search[i, 'lambda_l1'],
? ? ? ?lambda_l2 = grid_search[i, 'lambda_l2'],
? ? ? ?drop_rate = grid_search[i, 'drop_rate'],
? ? ? ?max_drop = grid_search[i, 'max_drop']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_learning_rate_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_learning_rate_1
ggplot(data = grid_search, aes(x = learning_rate, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
結(jié)論：learning_rate=.11時，auc最大。
3.調(diào)試num_leaves參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .11,
? ?num_leaves = seq(100, 800, 50),
? ?max_bin=1000,
? ?min_data_in_bin = 8,
? ?feature_fraction = .62,
? ?min_sum_hessian = .005,
? ?lambda_l1 = 0,
? ?lambda_l2 = 0,
? ?drop_rate = .2,
? ?max_drop = 5
)
perf_num_leaves_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 7 + 1) / sum(lgb_tr$TARGET * 7 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves'],
? ? ? ?max_bin = grid_search[i, 'max_bin'],
? ? ? ?min_data_in_bin = grid_search[i, 'min_data_in_bin'],
? ? ? ?feature_fraction = grid_search[i, 'feature_fraction'],
? ? ? ?min_sum_hessian = grid_search[i, 'min_sum_hessian'],
? ? ? ?lambda_l1 = grid_search[i, 'lambda_l1'],
? ? ? ?lambda_l2 = grid_search[i, 'lambda_l2'],
? ? ? ?drop_rate = grid_search[i, 'drop_rate'],
? ? ? ?max_drop = grid_search[i, 'max_drop']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_num_leaves_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_num_leaves_1
ggplot(data = grid_search, aes(x = num_leaves, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
結(jié)論：num_leaves=200時，auc最大。
4.調(diào)試max_bin參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .11,
? ?num_leaves = 200,
? ?max_bin = seq(100, 1500, 100),
? ?min_data_in_bin = 8,
? ?feature_fraction = .62,
? ?min_sum_hessian = .005,
? ?lambda_l1 = 0,
? ?lambda_l2 = 0,
? ?drop_rate = .2,
? ?max_drop = 5
)
perf_max_bin_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 7 + 1) / sum(lgb_tr$TARGET * 7 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves'],
? ? ? ?max_bin = grid_search[i, 'max_bin'],
? ? ? ?min_data_in_bin = grid_search[i, 'min_data_in_bin'],
? ? ? ?feature_fraction = grid_search[i, 'feature_fraction'],
? ? ? ?min_sum_hessian = grid_search[i, 'min_sum_hessian'],
? ? ? ?lambda_l1 = grid_search[i, 'lambda_l1'],
? ? ? ?lambda_l2 = grid_search[i, 'lambda_l2'],
? ? ? ?drop_rate = grid_search[i, 'drop_rate'],
? ? ? ?max_drop = grid_search[i, 'max_drop']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_max_bin_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_max_bin_1
ggplot(data = grid_search, aes(x = max_bin, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
結(jié)論：max_bin=600時，auc最大;400，800也是可接受值。
5.調(diào)試min_data_in_bin參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .11,
? ?num_leaves = 200,
? ?max_bin = 600,
? ?min_data_in_bin = seq(5, 50, 5),
? ?feature_fraction = .62,
? ?min_sum_hessian = .005,
? ?lambda_l1 = 0,
? ?lambda_l2 = 0,
? ?drop_rate = .2,
? ?max_drop = 5
)
perf_min_data_in_bin_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 7 + 1) / sum(lgb_tr$TARGET * 7 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves'],
? ? ? ?max_bin = grid_search[i, 'max_bin'],
? ? ? ?min_data_in_bin = grid_search[i, 'min_data_in_bin'],
? ? ? ?feature_fraction = grid_search[i, 'feature_fraction'],
? ? ? ?min_sum_hessian = grid_search[i, 'min_sum_hessian'],
? ? ? ?lambda_l1 = grid_search[i, 'lambda_l1'],
? ? ? ?lambda_l2 = grid_search[i, 'lambda_l2'],
? ? ? ?drop_rate = grid_search[i, 'drop_rate'],
? ? ? ?max_drop = grid_search[i, 'max_drop']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_min_data_in_bin_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_min_data_in_bin_1
ggplot(data = grid_search, aes(x = min_data_in_bin, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
結(jié)論：min_data_in_bin=45時，auc最大；其中25是可接受值。
6.調(diào)試feature_fraction參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .11,
? ?num_leaves = 200,
? ?max_bin = 600,
? ?min_data_in_bin = 45,
? ?feature_fraction = seq(.5, .9, .02),
? ?min_sum_hessian = .005,
? ?lambda_l1 = 0,
? ?lambda_l2 = 0,
? ?drop_rate = .2,
? ?max_drop = 5
)
perf_feature_fraction_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 7 + 1) / sum(lgb_tr$TARGET * 7 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves'],
? ? ? ?max_bin = grid_search[i, 'max_bin'],
? ? ? ?min_data_in_bin = grid_search[i, 'min_data_in_bin'],
? ? ? ?feature_fraction = grid_search[i, 'feature_fraction'],
? ? ? ?min_sum_hessian = grid_search[i, 'min_sum_hessian'],
? ? ? ?lambda_l1 = grid_search[i, 'lambda_l1'],
? ? ? ?lambda_l2 = grid_search[i, 'lambda_l2'],
? ? ? ?drop_rate = grid_search[i, 'drop_rate'],
? ? ? ?max_drop = grid_search[i, 'max_drop']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_feature_fraction_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_feature_fraction_1
ggplot(data = grid_search, aes(x = feature_fraction, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
結(jié)論：feature_fraction=.54時，auc最大, .56, .58時也較好。
7.調(diào)試min_sum_hessian參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .11,
? ?num_leaves = 200,
? ?max_bin = 600,
? ?min_data_in_bin = 45,
? ?feature_fraction = .54,
? ?min_sum_hessian = seq(.001, .008, .0005),
? ?lambda_l1 = 0,
? ?lambda_l2 = 0,
? ?drop_rate = .2,
? ?max_drop = 5
)
perf_min_sum_hessian_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 7 + 1) / sum(lgb_tr$TARGET * 7 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves'],
? ? ? ?max_bin = grid_search[i, 'max_bin'],
? ? ? ?min_data_in_bin = grid_search[i, 'min_data_in_bin'],
? ? ? ?feature_fraction = grid_search[i, 'feature_fraction'],
? ? ? ?min_sum_hessian = grid_search[i, 'min_sum_hessian'],
? ? ? ?lambda_l1 = grid_search[i, 'lambda_l1'],
? ? ? ?lambda_l2 = grid_search[i, 'lambda_l2'],
? ? ? ?drop_rate = grid_search[i, 'drop_rate'],
? ? ? ?max_drop = grid_search[i, 'max_drop']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_min_sum_hessian_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_min_sum_hessian_1
ggplot(data = grid_search, aes(x = min_sum_hessian, y = perf)) +
? ?geom_point() +
? ?geom_smooth()
結(jié)論：min_sum_hessian=0.0065時auc取得最大值，取min_sum_hessian=0.003，0.0055時可接受。
8.調(diào)試lambda參數(shù)
grid_search <- expand.grid(
? ?learning_rate = .11,
? ?num_leaves = 200,
? ?max_bin = 600,
? ?min_data_in_bin = 45,
? ?feature_fraction = .54,
? ?min_sum_hessian = 0.0065,
? ?lambda_l1 = seq(0, .001, .0002),
? ?lambda_l2 = seq(0, .001, .0002),
? ?drop_rate = .2,
? ?max_drop = 5
)
perf_lambda_1 <- numeric(length = nrow(grid_search))
for(i in 1:nrow(grid_search)){
? ?lgb_weight <- (lgb_tr$TARGET * 7 + 1) / sum(lgb_tr$TARGET * 7 + 1)
? ?
? ?lgb_train <- lgb.Dataset(
? ? ? ?data = data.matrix(lgb_tr[, 1:148]),
? ? ? ?label = lgb_tr$TARGET,
? ? ? ?free_raw_data = FALSE,
? ? ? ?weight = lgb_weight
? ?)
? ?
? ?# 參數(shù)
? ?params <- list(
? ? ? ?objective = 'binary',
? ? ? ?metric = 'auc',
? ? ? ?learning_rate = grid_search[i, 'learning_rate'],
? ? ? ?num_leaves = grid_search[i, 'num_leaves'],
? ? ? ?max_bin = grid_search[i, 'max_bin'],
? ? ? ?min_data_in_bin = grid_search[i, 'min_data_in_bin'],
? ? ? ?feature_fraction = grid_search[i, 'feature_fraction'],
? ? ? ?min_sum_hessian = grid_search[i, 'min_sum_hessian'],
? ? ? ?lambda_l1 = grid_search[i, 'lambda_l1'],
? ? ? ?lambda_l2 = grid_search[i, 'lambda_l2'],
? ? ? ?drop_rate = grid_search[i, 'drop_rate'],
? ? ? ?max_drop = grid_search[i, 'max_drop']
? ?)
? ?# 交叉驗證
? ?lgb_tr_mod <- lgb.cv(
? ? ? ?params,
? ? ? ?data = lgb_train,
? ? ? ?nrounds = 300,
? ? ? ?stratified = TRUE,
? ? ? ?nfold = 10,
? ? ? ?num_threads = 2,
? ? ? ?early_stopping_rounds = 10
? ?)
? ?perf_lambda_1[i] <- unlist(lgb_tr_mod$record_evals$valid$auc$eval)[length(unlist(lgb_tr_mod$record_evals$valid$auc$eval))]
}
grid_search$perf <- perf_lambda_1
ggplot(data = grid_search, aes(x = lambda_l1, y = perf)) +
? ?geom_point() +
? ?facet_wrap(~ lambda_l2, nrow = 5)
結(jié)論：lambda與auc整體呈負相關(guān)，取lambda_l1=.0002, lambda_l2 = .0004
9.調(diào)試drop_rate參數(shù)
結(jié)論：drop_rate=.4時取到最大值，.15, .25可接受。
10.調(diào)試max_drop參數(shù)
結(jié)論：drop_rate=.4時取到最大值，.15, .25可接受。
預測
1.權(quán)重
lgb_weight <- (lgb_tr$TARGET * 7 + 1) / sum(lgb_tr$TARGET * 7 + 1)
2.訓練數(shù)據(jù)集
lgb_train <- lgb.Dataset(
? ?data = data.matrix(lgb_tr[, 1:148]),
? ?label = lgb_tr$TARGET,
? ?free_raw_data = FALSE,
? ?weight = lgb_weight
)
3.訓練
# 參數(shù)
params <- list(
? ?learning_rate = .11,
? ?num_leaves = 200,
? ?max_bin = 600,
? ?min_data_in_bin = 45,
? ?feature_fraction = .54,
? ?min_sum_hessian = 0.0065,
? ?lambda_l1 = .0002,
? ?lambda_l2 = .0004,
? ?drop_rate = .4,
? ?max_drop = 14
)
# 模型
lgb_mod <- lightgbm(
? ?params = params,
? ?data = lgb_train,
? ?nrounds = 300,
? ?early_stopping_rounds = 10,
? ?num_threads = 2
)
# 預測
lgb.pred <- predict(lgb_mod, data.matrix(lgb_te))
4.結(jié)果
lgb.pred2 <- matrix(unlist(lgb.pred), ncol = 1)
lgb.pred3 <- data.frame(lgb.pred2)
5.輸出
write.csv(lgb.pred3, "C:/Users/Administrator/Documents/kaggle/scs_lgb/lgb.pred1_tr.csv")
注：此處給在校讀書的朋友一些建議：
1.在學校學習機器學習算法時，測試所用數(shù)據(jù)量一般較少，因此可以嘗試大多數(shù)算法，大多數(shù)的R函數(shù)，例如測試隨機森林算法時，可以選擇randomforest包，如果數(shù)據(jù)量稍微增多，可以設(shè)置并行運算，但是如果數(shù)據(jù)量達到GB級別，并行運算randomforest包也處理不了了，并且內(nèi)存會溢出；建議使用專業(yè)版R中的函數(shù)；
2.學校學習主要針對理論進行學習，測試數(shù)據(jù)一般較為干凈，實際數(shù)據(jù)結(jié)構(gòu)一般更為復雜一些。
往期精彩：
【原創(chuàng)首發(fā)】機器學習公式推導與代碼實現(xiàn)30講.pdf
【原創(chuàng)首發(fā)】深度學習語義分割理論與實戰(zhàn)指南.pdf