Author: Jiachang Liu
Date: Last Compiled on November 09, 2022
install.packages("reticulate", repos = "http://cran.us.r-project.org")
library(reticulate)version <- "3.9.12"
install_python(version)
virtualenv_create(envname="FasterRisk-environment", version = version)
use_virtualenv("FasterRisk-environment")py_install("fasterrisk", pip=TRUE, envname="FasterRisk-environment")fasterrisk <- import("fasterrisk")data_dir_path <- "../tests"
train_data_file_path <- "../tests/adult_train_data.csv"
test_data_file_path <- "../tests/adult_test_data.csv"
if (!file.exists(data_dir_path)){
# create the data folder if it doesn't exist
dir.create(file.path(data_dir_path))
}
if (!file.exists(train_data_file_path)){
# download sample training data if it doesn't exists
fasterrisk$utils$download_file_from_google_drive('1nuWn0QVG8tk3AN4I4f3abWLcFEP3WPec', train_data_file_path)
}
if (!file.exists(train_data_file_path)){
# download sample test data if it doesn't exists
fasterrisk$utils$download_file_from_google_drive('1TyBO02LiGfHbatPWU4nzc8AndtIF-7WH', test_data_file_path)
}np <- import("numpy", convert=FALSE)
train_df <- read.csv(train_data_file_path)
train_data <- data.matrix(train_df)
X_train <- np$array(train_data[, 2:ncol(train_data)])
y_train <- np$array(train_data[, 1], dtype=np$int) # Check your data! y label must be +1/-1.
test_df <- read.csv(test_data_file_path)
test_data <- data.matrix(test_df)
X_test <- np$array(test_data[, 2:ncol(test_data)])
y_test <- np$array(test_data[, 1], dtype=np$int) # Check your data! y label must be +1/-1.sparsity <- as.integer(5)
parent_size <- as.integer(10)
RiskScoreOptimizer_m <- fasterrisk$fasterrisk$RiskScoreOptimizer(X = X_train, y = y_train, k = sparsity, parent_size = parent_size)start_time <- Sys.time()
RiskScoreOptimizer_m$optimize()
sprintf("Optimization takes %f seconds.", Sys.time() - start_time)## [1] "Optimization takes 13.313264 seconds."
solutions = RiskScoreOptimizer_m$get_models()
multipliers = solutions[1][[1]]
sparseDiversePool_beta0_integer = solutions[2][[1]]
sparseDiversePool_betas_integer = solutions[3][[1]]
sprintf("We generate %d risk score models from the sparse diverse pool", length(multipliers))## [1] "We generate 50 risk score models from the sparse diverse pool"
model_index = 1 # first model
multiplier = multipliers[model_index]
intercept = sparseDiversePool_beta0_integer[model_index]
coefficients = np$array(sparseDiversePool_betas_integer[model_index, ]) # each row of sparseDiversePool_betas_integer corresponds to a solution of coefficientsRiskScoreClassifier_m = fasterrisk$fasterrisk$RiskScoreClassifier(multiplier, intercept, coefficients)y_test_pred = RiskScoreClassifier_m$predict(X_test)
print("y_test are predicted to be (first 10 values):")## [1] "y_test are predicted to be (first 10 values):"
y_test_pred[1:10]## [1] -1 -1 -1 -1 -1 -1 -1 1 -1 -1
y_test_pred_prob = RiskScoreClassifier_m$predict_prob(X_test)
print("The risk probabilities of having y_test to be +1 are (first 10 values):")## [1] "The risk probabilities of having y_test to be +1 are (first 10 values):"
y_test_pred_prob[1:10]## [1] 0.13308868 0.34872682 0.34872682 0.04216029 0.13308868 0.34872682
## [7] 0.04216029 0.65127318 0.34872682 0.01246260
X_featureNames = list(colnames(train_df)[-1])[[1]] # extract the column names from the dataframe. Feature names exclude the first column name because it is the name for the y label
RiskScoreClassifier_m$reset_featureNames(X_featureNames)
tmp_str = py_capture_output(RiskScoreClassifier_m$print_model_card(), type = c("stdout", "stderr")) # capture the python standard output and print it inside R
cat(tmp_str)## The Risk Score is:
## 1. Age_22_to_29 -2 point(s) | ...
## 2. HSDiploma -2 point(s) | + ...
## 3. NoHS -4 point(s) | + ...
## 4. Married 4 point(s) | + ...
## 5. AnyCapitalGains 3 point(s) | + ...
## SCORE | =
## SCORE | -8.0 | -6.0 | -5.0 | -4.0 | -3.0 | -2.0 | -1.0 |
## RISK | 0.1% | 0.4% | 0.7% | 1.2% | 2.3% | 4.2% | 7.6% |
## SCORE | 0.0 | 1.0 | 2.0 | 3.0 | 4.0 | 5.0 | 7.0 |
## RISK | 13.3% | 22.3% | 34.9% | 50.0% | 65.1% | 77.7% | 92.4% |
num_models = min(10, length(multipliers))
for (model_index in 1:num_models){
multiplier = multipliers[model_index]
intercept = sparseDiversePool_beta0_integer[model_index]
coefficients = np$array(sparseDiversePool_betas_integer[model_index, ]) # each row of sparseDiversePool_betas_integer corresponds to a solution of coefficients
RiskScoreClassifier_m = fasterrisk$fasterrisk$RiskScoreClassifier(multiplier, intercept, coefficients)
RiskScoreClassifier_m$reset_featureNames(X_featureNames)
tmp_str = py_capture_output(RiskScoreClassifier_m$print_model_card(), type = c("stdout", "stderr")) # capture the python standard output and print it inside R
cat(tmp_str)
train_loss = RiskScoreClassifier_m$compute_logisticLoss(X_train, y_train)
train_results = RiskScoreClassifier_m$get_acc_and_auc(X_train, y_train)
train_acc = train_results[1][[1]]
train_auc = train_results[2][[1]]
test_results = RiskScoreClassifier_m$get_acc_and_auc(X_test, y_test)
test_acc = test_results[1][[1]]
test_auc = test_results[2][[1]]
tmp_str = sprintf("The logistic loss on the training set is %f", train_loss)
print(tmp_str)
tmp_str = sprintf("The training accuracy and AUC are %f and %f", train_acc*100, train_auc)
print(tmp_str)
tmp_str = sprintf("The test accuracy and AUC are are %f and %f", test_acc*100, test_auc)
print(tmp_str)
cat("\n")
}## The Risk Score is:
## 1. Age_22_to_29 -2 point(s) | ...
## 2. HSDiploma -2 point(s) | + ...
## 3. NoHS -4 point(s) | + ...
## 4. Married 4 point(s) | + ...
## 5. AnyCapitalGains 3 point(s) | + ...
## SCORE | =
## SCORE | -8.0 | -6.0 | -5.0 | -4.0 | -3.0 | -2.0 | -1.0 |
## RISK | 0.1% | 0.4% | 0.7% | 1.2% | 2.3% | 4.2% | 7.6% |
## SCORE | 0.0 | 1.0 | 2.0 | 3.0 | 4.0 | 5.0 | 7.0 |
## RISK | 13.3% | 22.3% | 34.9% | 50.0% | 65.1% | 77.7% | 92.4% |
##
## [1] "The logistic loss on the training set is 9798.652347"
## [1] "The training accuracy and AUC are 82.575147 and 0.861817"
## [1] "The test accuracy and AUC are are 81.787469 and 0.856367"
##
## The Risk Score is:
## 1. HSDiploma -2 point(s) | ...
## 2. NoHS -4 point(s) | + ...
## 3. Married 4 point(s) | + ...
## 4. WorkHrsPerWeek_lt_40 -2 point(s) | + ...
## 5. AnyCapitalGains 3 point(s) | + ...
## SCORE | =
## SCORE | -8.0 | -6.0 | -5.0 | -4.0 | -3.0 | -2.0 | -1.0 |
## RISK | 0.1% | 0.4% | 0.7% | 1.3% | 2.5% | 4.4% | 7.9% |
## SCORE | 0.0 | 1.0 | 2.0 | 3.0 | 4.0 | 5.0 | 7.0 |
## RISK | 13.7% | 22.7% | 35.1% | 50.0% | 64.9% | 77.3% | 92.1% |
##
## [1] "The logistic loss on the training set is 9859.615758"
## [1] "The training accuracy and AUC are 82.333295 and 0.859661"
## [1] "The test accuracy and AUC are are 81.848894 and 0.853752"
##
## The Risk Score is:
## 1. HSDiploma -3 point(s) | ...
## 2. NoHS -5 point(s) | + ...
## 3. JobManagerial 2 point(s) | + ...
## 4. Married 5 point(s) | + ...
## 5. AnyCapitalGains 3 point(s) | + ...
## SCORE | =
## SCORE | -8.0 | -6.0 | -5.0 | -3.0 | -2.0 | -1.0 | 0.0 |
## RISK | 0.2% | 0.6% | 1.0% | 2.7% | 4.4% | 7.2% | 11.4% |
## SCORE | 2.0 | 3.0 | 4.0 | 5.0 | 7.0 | 8.0 | 10.0 |
## RISK | 26.4% | 37.5% | 50.0% | 62.5% | 82.3% | 88.6% | 95.6% |
##
## [1] "The logistic loss on the training set is 9883.324462"
## [1] "The training accuracy and AUC are 82.268033 and 0.859570"
## [1] "The test accuracy and AUC are are 81.511057 and 0.854237"
##
## The Risk Score is:
## 1. HSDiploma -3 point(s) | ...
## 2. NoHS -5 point(s) | + ...
## 3. Married 5 point(s) | + ...
## 4. WorkHrsPerWeek_geq_50 1 point(s) | + ...
## 5. AnyCapitalGains 3 point(s) | + ...
## SCORE | =
## SCORE | -8.0 | -7.0 | -5.0 | -4.0 | -3.0 | -2.0 | -1.0 | 0.0 |
## RISK | 0.2% | 0.3% | 0.9% | 1.5% | 2.5% | 4.1% | 6.8% | 10.9% |
## SCORE | 1.0 | 2.0 | 3.0 | 4.0 | 5.0 | 6.0 | 8.0 | 9.0 |
## RISK | 17.2% | 25.9% | 37.2% | 50.0% | 62.8% | 74.1% | 89.1% | 93.2% |
##
## [1] "The logistic loss on the training set is 9895.728068"
## [1] "The training accuracy and AUC are 82.179738 and 0.861227"
## [1] "The test accuracy and AUC are are 81.342138 and 0.856065"
##
## The Risk Score is:
## 1. Age_45_to_59 1 point(s) | ...
## 2. HSDiploma -2 point(s) | + ...
## 3. NoHS -5 point(s) | + ...
## 4. Married 4 point(s) | + ...
## 5. AnyCapitalGains 3 point(s) | + ...
## SCORE | =
## SCORE | -7.0 | -6.0 | -5.0 | -4.0 | -3.0 | -2.0 | -1.0 | 0.0 |
## RISK | 0.2% | 0.4% | 0.7% | 1.2% | 2.0% | 3.5% | 5.9% | 9.9% |
## SCORE | 1.0 | 2.0 | 3.0 | 4.0 | 5.0 | 6.0 | 7.0 | 8.0 |
## RISK | 16.0% | 24.9% | 36.5% | 50.0% | 63.5% | 75.1% | 84.0% | 90.1% |
##
## [1] "The logistic loss on the training set is 9914.759742"
## [1] "The training accuracy and AUC are 80.655687 and 0.862991"
## [1] "The test accuracy and AUC are are 80.052211 and 0.856397"
##
## The Risk Score is:
## 1. HSDiploma -3 point(s) | ...
## 2. NoHS -5 point(s) | + ...
## 3. Married 5 point(s) | + ...
## 4. AnyCapitalGains 3 point(s) | + ...
## 5. AnyCapitalLoss 2 point(s) | + ...
## SCORE | =
## SCORE | -8.0 | -6.0 | -5.0 | -3.0 | -2.0 | -1.0 | 0.0 |
## RISK | 0.2% | 0.5% | 0.8% | 2.3% | 3.9% | 6.4% | 10.5% |
## SCORE | 2.0 | 3.0 | 4.0 | 5.0 | 7.0 | 8.0 | 10.0 |
## RISK | 25.5% | 36.9% | 50.0% | 63.1% | 83.3% | 89.5% | 96.1% |
##
## [1] "The logistic loss on the training set is 9923.881690"
## [1] "The training accuracy and AUC are 82.179738 and 0.857475"
## [1] "The test accuracy and AUC are are 81.342138 and 0.852014"
##
## The Risk Score is:
## 1. HSDiploma -2 point(s) | ...
## 2. ProfVocOrAS -1 point(s) | + ...
## 3. NoHS -4 point(s) | + ...
## 4. Married 3 point(s) | + ...
## 5. AnyCapitalGains 2 point(s) | + ...
## SCORE | =
## SCORE | -7.0 | -6.0 | -5.0 | -4.0 | -3.0 | -2.0 | -1.0 |
## RISK | 0.1% | 0.1% | 0.3% | 0.8% | 1.7% | 3.7% | 8.0% |
## SCORE | 0.0 | 1.0 | 2.0 | 3.0 | 4.0 | 5.0 |
## RISK | 16.4% | 30.7% | 50.0% | 69.3% | 83.6% | 92.0% |
##
## [1] "The logistic loss on the training set is 9980.639484"
## [1] "The training accuracy and AUC are 82.172060 and 0.855975"
## [1] "The test accuracy and AUC are are 81.234644 and 0.849403"
##
## The Risk Score is:
## 1. HSDiploma -2 point(s) | ...
## 2. NoHS -4 point(s) | + ...
## 3. Married 3 point(s) | + ...
## 4. NeverMarried -1 point(s) | + ...
## 5. AnyCapitalGains 2 point(s) | + ...
## SCORE | =
## SCORE | -7.0 | -6.0 | -5.0 | -4.0 | -3.0 | -2.0 | -1.0 |
## RISK | 0.1% | 0.3% | 0.6% | 1.2% | 2.4% | 4.9% | 9.8% |
## SCORE | 0.0 | 1.0 | 2.0 | 3.0 | 4.0 | 5.0 |
## RISK | 18.6% | 32.3% | 50.0% | 67.7% | 81.4% | 90.2% |
##
## [1] "The logistic loss on the training set is 9988.041002"
## [1] "The training accuracy and AUC are 82.179738 and 0.855476"
## [1] "The test accuracy and AUC are are 81.342138 and 0.849000"
##
## The Risk Score is:
## 1. HSDiploma -2 point(s) | ...
## 2. NoHS -4 point(s) | + ...
## 3. Married 4 point(s) | + ...
## 4. DivorcedOrSeparated 1 point(s) | + ...
## 5. AnyCapitalGains 2 point(s) | + ...
## SCORE | =
## SCORE | -6.0 | -5.0 | -4.0 | -3.0 | -2.0 | -1.0 | 0.0 |
## RISK | 0.1% | 0.3% | 0.6% | 1.2% | 2.6% | 5.1% | 10.1% |
## SCORE | 1.0 | 2.0 | 3.0 | 4.0 | 5.0 | 6.0 | 7.0 |
## RISK | 18.9% | 32.6% | 50.0% | 67.4% | 81.1% | 89.9% | 94.9% |
##
## [1] "The logistic loss on the training set is 10000.803139"
## [1] "The training accuracy and AUC are 82.179738 and 0.854825"
## [1] "The test accuracy and AUC are are 81.342138 and 0.848115"
##
## The Risk Score is:
## 1. HSDiploma -2 point(s) | ...
## 2. NoHS -4 point(s) | + ...
## 3. JobService -1 point(s) | + ...
## 4. Married 4 point(s) | + ...
## 5. AnyCapitalGains 3 point(s) | + ...
## SCORE | =
## SCORE | -7.0 | -6.0 | -5.0 | -4.0 | -3.0 | -2.0 | -1.0 | 0.0 |
## RISK | 0.2% | 0.4% | 0.7% | 1.3% | 2.4% | 4.3% | 7.7% | 13.4% |
## SCORE | 1.0 | 2.0 | 3.0 | 4.0 | 5.0 | 6.0 | 7.0 |
## RISK | 22.4% | 35.0% | 50.0% | 65.0% | 77.6% | 86.6% | 92.3% |
##
## [1] "The logistic loss on the training set is 10000.838407"
## [1] "The training accuracy and AUC are 82.175899 and 0.857572"
## [1] "The test accuracy and AUC are are 81.464988 and 0.851669"