Deep Learning

---

0.1 Deep Learning (Neural Networks) _Supervised

H2O 를 활용한 Deep Learning (Neural Networks)

[참조 1] http://docs.h2o.ai/h2o/latest-stable/h2o-docs/data-science/deep-learning.html#deep-learning-neural-networks

---

0.1.1 packages

library(h2o)

## Warning: package 'h2o' was built under R version 4.0.3

# 초기 준비
h2o.init()

##  Connection successful!
## 
## R is connected to the H2O cluster: 
##     H2O cluster uptime:         5 hours 11 minutes 
##     H2O cluster timezone:       Asia/Seoul 
##     H2O data parsing timezone:  UTC 
##     H2O cluster version:        3.32.0.1 
##     H2O cluster version age:    25 days  
##     H2O cluster name:           H2O_started_from_R_user_uho906 
##     H2O cluster total nodes:    1 
##     H2O cluster total memory:   3.96 GB 
##     H2O cluster total cores:    4 
##     H2O cluster allowed cores:  4 
##     H2O cluster healthy:        TRUE 
##     H2O Connection ip:          localhost 
##     H2O Connection port:        54321 
##     H2O Connection proxy:       NA 
##     H2O Internal Security:      FALSE 
##     H2O API Extensions:         Amazon S3, Algos, AutoML, Core V3, TargetEncoder, Core V4 
##     R Version:                  R version 4.0.2 (2020-06-22)

---

0.1.2 data import

# Import the insurance dataset into H2O:
insurance <- h2o.importFile("http://s3.amazonaws.com/h2o-public-test-data/smalldata/glm_test/insurance.csv")

## 
  |                                                                            
  |                                                                      |   0%
  |                                                                            
  |======================================================================| 100%

insurance

##   District  Group   Age Holders Claims
## 1        1    <1l   <25     197     38
## 2        1    <1l 25-29     264     35
## 3        1    <1l 30-35     246     20
## 4        1    <1l   >35    1680    156
## 5        1 1-1.5l   <25     284     63
## 6        1 1-1.5l 25-29     536     84
## 
## [64 rows x 5 columns]

---

0.1.3 data mumming

# Set the factors:
offset = log(insurance$Holders)
insurance$Holders <- as.factor(insurance$Holders)
insurance$Age <- as.factor(insurance$Age)
insurance$Group <- as.factor(insurance$Group)
insurance$District <- as.factor(insurance$District)

---

0.1.4 modeling

# Build and train the model:
dl <- h2o.deeplearning(x = 1:3,
                       y = "Claims",
                       distribution = "tweedie",
                       hidden = c(1),
                       epochs = 1000,
                       train_samples_per_iteration = -1,
                       reproducible = TRUE,
                       activation = "Tanh",
                       single_node_mode = FALSE,
                       balance_classes = FALSE,
                       force_load_balance = FALSE,
                       seed = 23123,
                       tweedie_power = 1.5,
                       score_training_samples = 0,
                       score_validation_samples = 0,
                       training_frame = insurance,
                       stopping_rounds = 0)

## 
  |                                                                            
  |                                                                      |   0%
  |                                                                            
  |=                                                                     |   1%
  |                                                                            
  |===========                                                           |  16%
  |                                                                            
  |===================                                                   |  27%
  |                                                                            
  |=========================                                             |  36%
  |                                                                            
  |===============================                                       |  44%
  |                                                                            
  |====================================                                  |  51%
  |                                                                            
  |========================================                              |  58%
  |                                                                            
  |=============================================                         |  64%
  |                                                                            
  |==================================================                    |  71%
  |                                                                            
  |======================================================                |  78%
  |                                                                            
  |===========================================================           |  84%
  |                                                                            
  |===============================================================       |  90%
  |                                                                            
  |===================================================================   |  96%
  |                                                                            
  |======================================================================| 100%

# Eval performance:
perf <- h2o.performance(dl)
perf

## H2ORegressionMetrics: deeplearning
## ** Reported on training data. **
## ** Metrics reported on full training frame **
## 
## MSE:  373.5344
## RMSE:  19.32704
## MAE:  9.058789
## RMSLE:  0.3281787
## Mean Residual Deviance :  0.5560917

---

0.1.5 prediction

# Generate predictions on a test set (if necessary):
pred <- h2o.predict(dl, newdata = insurance)

## 
  |                                                                            
  |                                                                      |   0%
  |                                                                            
  |======================================================================| 100%

pred

##     predict
## 1  17.46302
## 2  33.08835
## 3  38.73943
## 4 166.64276
## 5  50.03000
## 6  86.78265
## 
## [64 rows x 1 column]

---