Project: NASA - Nearest Earth Objects
Amir Kavie
2022-07-26
Executive Summary
In this project an in depth analysis of the latest database on Nearest Earth Objects has been performed.
Although the Accuracy of our Logistic Regression was 90.4%, we had a high rate of false positives which was reprented in the form of high Sensitivity of 98.9%, low Specificity 8.5%, and an almost unacceptably low ROC AUC of 11.9%.
Through the implementation of a Random Forest, we observe that the most important variables are “miss_distance” and “relative_velocity”, which are followed by “est_diameter_max”, “absolute_magnitude” and “est_diameter_min”.
Our Random Forest has an Accuracy of 91.8%, a Sensitivity of 97.8%, and a Specificity of 34%, which means that Random Forest is a more reliable model compared to the results of our Logistic Regression model.
Finally, by Boosting our Random Forest and tuning it, we find a more reliable model with an ROC AUC of 91.6%.
Data Import
During the data import process, I dropped the ‘id’ and ‘name’ columns, since they are man-made variables and have no relevance to our analysis.
library(ggplot2)
library(tidymodels)
library(dplyr)
library(xgboost)
library(yardstick)nasa_df <- read.csv("neo_v2.csv")
nasa_df$hazardous <- as.factor(nasa_df$hazardous)
# Converting KM to Meters
nasa_df$est_diameter_min <- nasa_df$est_diameter_min * 1000
nasa_df$est_diameter_max <- nasa_df$est_diameter_max * 1000
#nasa_df$orbiting_body <- as.factor(nasa_df$orbiting_body)
#nasa_df$sentry_object <- as.factor(nasa_df$sentry_object)
#nasa_df
#nrow(nasa_df)
#head(nasa_df)
nasa_df$id <- NULL
nasa_df$name <- NULL
nasa_df$orbiting_body <- NULL
nasa_df$sentry_object <- NULL
str(nasa_df)## 'data.frame': 90836 obs. of 6 variables:
## $ est_diameter_min : num 1198.3 265.8 722 96.5 255 ...
## $ est_diameter_max : num 2679 594 1615 216 570 ...
## $ relative_velocity : num 13569 73589 114259 24764 42738 ...
## $ miss_distance : num 54839744 61438127 49798725 25434973 46275567 ...
## $ absolute_magnitude: num 16.7 20 17.8 22.2 20.1 ...
## $ hazardous : Factor w/ 2 levels "False","True": 1 2 1 1 2 1 1 1 1 1 ...summary(nasa_df)## est_diameter_min est_diameter_max relative_velocity miss_distance
## Min. : 0.61 Min. : 1.36 Min. : 203.3 Min. : 6746
## 1st Qu.: 19.26 1st Qu.: 43.06 1st Qu.: 28619.0 1st Qu.:17210820
## Median : 48.37 Median : 108.15 Median : 44190.1 Median :37846579
## Mean : 127.43 Mean : 284.95 Mean : 48066.9 Mean :37066546
## 3rd Qu.: 143.40 3rd Qu.: 320.66 3rd Qu.: 62923.6 3rd Qu.:56548996
## Max. :37892.65 Max. :84730.54 Max. :236990.1 Max. :74798651
## absolute_magnitude hazardous
## Min. : 9.23 False:81996
## 1st Qu.:21.34 True : 8840
## Median :23.70
## Mean :23.53
## 3rd Qu.:25.70
## Max. :33.20Univariate Analysis
hazardous
ggplot(nasa_df, aes(hazardous)) +
geom_bar(color = "darkgreen", fill = "lightgreen", width = 0.2) +
theme_classic()
summary(nasa_df$hazardous)## False True
## 81996 8840est_diameter_min
ggplot(nasa_df, aes((est_diameter_min))) +
geom_histogram(color = "darkgreen", fill = "lightgreen") +
scale_x_log10() +
theme_classic()## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
summary(nasa_df$est_diameter_min)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.61 19.26 48.37 127.43 143.40 37892.65mean(nasa_df$est_diameter_min)## [1] 127.4321est_diameter_max
ggplot(nasa_df, aes(est_diameter_max)) +
geom_histogram(color = "darkgreen", fill = "lightgreen") +
scale_x_log10() +
theme_classic()## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
summary(nasa_df$est_diameter_max)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 1.36 43.06 108.15 284.95 320.66 84730.54relative_velocity
ggplot(nasa_df, aes(relative_velocity)) +
geom_histogram(color = "darkgreen", fill = "lightgreen") +
theme_classic()## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
summary(nasa_df$relative_velocity)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 203.3 28619.0 44190.1 48066.9 62923.6 236990.1miss_distance
ggplot(nasa_df, aes(miss_distance)) +
geom_histogram(color = "darkgreen", fill = "lightgreen") +
theme_classic()## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
summary(nasa_df$miss_distance)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 6746 17210820 37846579 37066546 56548996 74798651absolute_magnitude
ggplot(nasa_df, aes(absolute_magnitude)) +
geom_histogram(color = "darkgreen", fill = "lightgreen") +
theme_classic()## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.
summary(nasa_df$absolute_magnitude)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 9.23 21.34 23.70 23.53 25.70 33.20Bivariate Analysis
Hazardous vs. est_diameter_min
glm_mdl <- glm(hazardous ~ est_diameter_min, data = nasa_df, family = binomial)## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurredsummary(glm_mdl)##
## Call:
## glm(formula = hazardous ~ est_diameter_min, family = binomial,
## data = nasa_df)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -8.4904 -0.4230 -0.4000 -0.3922 2.2207
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -2.552e+00 1.369e-02 -186.35 <2e-16 ***
## est_diameter_min 1.985e-03 4.026e-05 49.31 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 57981 on 90835 degrees of freedom
## Residual deviance: 55084 on 90834 degrees of freedom
## AIC: 55088
##
## Number of Fisher Scoring iterations: 6coefficients(glm_mdl)## (Intercept) est_diameter_min
## -2.551719401 0.001985131ggplot(nasa_df, aes(est_diameter_min, as.numeric(hazardous) - 1)) +
geom_point() +
geom_smooth(method="glm", color="blue", se=FALSE,
method.args = list(family='binomial'))## `geom_smooth()` using formula 'y ~ x'## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
Hazardous vs. est_diameter_max
glm_mdl <- glm(hazardous ~ est_diameter_max, data = nasa_df, family = binomial)## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurredsummary(glm_mdl)##
## Call:
## glm(formula = hazardous ~ est_diameter_max, family = binomial,
## data = nasa_df)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -8.4904 -0.4230 -0.4000 -0.3922 2.2207
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -2.5517194 0.0136931 -186.35 <2e-16 ***
## est_diameter_max 0.0008878 0.0000180 49.31 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 57981 on 90835 degrees of freedom
## Residual deviance: 55084 on 90834 degrees of freedom
## AIC: 55088
##
## Number of Fisher Scoring iterations: 6coefficients(glm_mdl)## (Intercept) est_diameter_max
## -2.5517194008 0.0008877777ggplot(nasa_df, aes(est_diameter_max, as.numeric(hazardous) - 1)) +
geom_point() +
geom_smooth(method="glm", color="blue", se=FALSE,
method.args = list(family='binomial'))## `geom_smooth()` using formula 'y ~ x'## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred
Hazardous vs. relative_velocity
glm_mdl <- glm(hazardous ~ relative_velocity, data = nasa_df, family = binomial)
summary(glm_mdl)##
## Call:
## glm(formula = hazardous ~ relative_velocity, family = binomial,
## data = nasa_df)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.9757 -0.4755 -0.3868 -0.3267 2.5777
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -3.416e+00 2.646e-02 -129.10 <2e-16 ***
## relative_velocity 2.200e-05 3.987e-07 55.18 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 57981 on 90835 degrees of freedom
## Residual deviance: 55014 on 90834 degrees of freedom
## AIC: 55018
##
## Number of Fisher Scoring iterations: 5coefficients(glm_mdl)## (Intercept) relative_velocity
## -3.415572e+00 2.200084e-05ggplot(nasa_df, aes(relative_velocity, as.numeric(hazardous) - 1)) +
geom_point() +
geom_smooth(method="glm", color="blue", se=FALSE,
method.args = list(family='binomial'))## `geom_smooth()` using formula 'y ~ x'
Hazardous vs. miss_distance
glm_mdl <- glm(hazardous ~ miss_distance, data = nasa_df, family = binomial)
summary(glm_mdl)##
## Call:
## glm(formula = hazardous ~ miss_distance, family = binomial, data = nasa_df)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -0.5054 -0.4749 -0.4456 -0.4146 2.2600
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -2.474e+00 2.294e-02 -107.85 <2e-16 ***
## miss_distance 6.426e-09 5.049e-10 12.72 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 57981 on 90835 degrees of freedom
## Residual deviance: 57818 on 90834 degrees of freedom
## AIC: 57822
##
## Number of Fisher Scoring iterations: 5coefficients(glm_mdl)## (Intercept) miss_distance
## -2.473836e+00 6.425476e-09ggplot(nasa_df, aes(miss_distance, as.numeric(hazardous) - 1)) +
geom_point() +
geom_smooth(method="glm", color="blue", se=FALSE,
method.args = list(family='binomial'))## `geom_smooth()` using formula 'y ~ x'
Hazardous vs. absolute_magnitude
glm_mdl <- glm(hazardous ~ absolute_magnitude, data = nasa_df, family = binomial)
summary(glm_mdl)##
## Call:
## glm(formula = hazardous ~ absolute_magnitude, family = binomial,
## data = nasa_df)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.9540 -0.4217 -0.2430 -0.1423 2.2411
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) 9.099601 0.113777 79.98 <2e-16 ***
## absolute_magnitude -0.514568 0.005415 -95.02 <2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 57981 on 90835 degrees of freedom
## Residual deviance: 44949 on 90834 degrees of freedom
## AIC: 44953
##
## Number of Fisher Scoring iterations: 6coefficients(glm_mdl)## (Intercept) absolute_magnitude
## 9.0996008 -0.5145679ggplot(nasa_df, aes(absolute_magnitude, as.numeric(hazardous) - 1)) +
geom_point() +
geom_smooth(method="glm", color="blue", se=FALSE,
method.args = list(family='binomial'))## `geom_smooth()` using formula 'y ~ x'
Correlation Matrix
# Correlation
#cor(nasa_df$est_diameter_min, nasa_df$relative_velocity, method="pearson")
#nasa_df$hazardous <- as.numeric(nasa_df$hazardous)
#cor(nasa_df)library(corrplot)## Warning: package 'corrplot' was built under R version 4.2.1## corrplot 0.92 loaded#corrplot(nasa_df, method = "circle")Pairs Chart
pairs(data = nasa_df, hazardous ~ est_diameter_min + est_diameter_max +
relative_velocity + miss_distance + absolute_magnitude, col = nasa_df$hazardous)
Logistic Regression
# Create data split object
nasa_split <- initial_split(nasa_df, prop = 0.75, strata = hazardous)
# Create the training data
nasa_training <- nasa_split %>%
training()
# Create the test data
nasa_test <- nasa_split %>%
testing()
# Check the number of rows
nrow(nasa_training)## [1] 68127nrow(nasa_test)## [1] 22709# Specify a logistic regression model
logistic_model <- logistic_reg() %>%
# Set the engine
set_engine('glm') %>%
# Set the mode
set_mode('classification')
# Fit to training data
logistic_fit <- logistic_model %>%
fit(hazardous ~ est_diameter_min + est_diameter_max + relative_velocity +
miss_distance + absolute_magnitude,
data = nasa_training)## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred# Print model fit object
logistic_fit## parsnip model object
##
##
## Call: stats::glm(formula = hazardous ~ est_diameter_min + est_diameter_max +
## relative_velocity + miss_distance + absolute_magnitude, family = stats::binomial,
## data = data)
##
## Coefficients:
## (Intercept) est_diameter_min est_diameter_max relative_velocity
## 2.457e+01 -3.357e+06 1.501e+06 7.789e-06
## miss_distance absolute_magnitude
## -2.085e-08 -1.169e+00
##
## Degrees of Freedom: 68126 Total (i.e. Null); 68121 Residual
## Null Deviance: 43410
## Residual Deviance: 30260 AIC: 30270Making Predictions based on Logistic Regression
# Predict outcome categories
class_preds <- predict(logistic_fit, new_data = nasa_test,
type = 'class')
# Obtain estimated probabilities for each outcome value
prob_preds <- predict(logistic_fit, new_data = nasa_test,
type = 'prob')
# Combine test set results
nasa_results <- nasa_test %>%
select(hazardous) %>%
bind_cols(class_preds, prob_preds)
# View results tibble
#nasa_resultsConfusion Matrix
# Calculate the confusion matrix
head(nasa_results)## hazardous .pred_class .pred_False .pred_True
## 1 False False 0.6001969 3.998031e-01
## 2 False False 0.9199434 8.005656e-02
## 3 True False 0.7420153 2.579847e-01
## 4 False False 0.9999114 8.858862e-05
## 5 False False 0.8853639 1.146361e-01
## 6 False False 0.9998246 1.753648e-04conf_mat(nasa_results, truth = hazardous,
estimate = .pred_class)## Truth
## Prediction False True
## False 20269 2058
## True 214 168Accuracy
# Calculate the accuracy
accuracy(nasa_results, truth = hazardous,
estimate = .pred_class)## # A tibble: 1 × 3
## .metric .estimator .estimate
## <chr> <chr> <dbl>
## 1 accuracy binary 0.900Sensitivity
# Calculate the sensitivity
sens(nasa_results, truth = hazardous,
estimate = .pred_class)## # A tibble: 1 × 3
## .metric .estimator .estimate
## <chr> <chr> <dbl>
## 1 sens binary 0.990Specificity
# Calculate the specificity
spec(nasa_results, truth = hazardous,
estimate = .pred_class)## # A tibble: 1 × 3
## .metric .estimator .estimate
## <chr> <chr> <dbl>
## 1 spec binary 0.0755# Create a Heat Map of confusion matrix
conf_mat(nasa_results,
truth = hazardous,
estimate = .pred_class) %>%
# Create a heat map
autoplot(type = 'heatmap')
# Create a Mosaic Plor of confusion matrix
conf_mat(nasa_results,
truth = hazardous,
estimate = .pred_class) %>%
# Create a mosaic plot
autoplot(type = "mosaic")
# ROC & AUC
# Calculate metrics across thresholds
threshold_df <- nasa_results %>%
roc_curve(truth = hazardous, .pred_True)
# View results
threshold_df## # A tibble: 22,710 × 3
## .threshold specificity sensitivity
## <dbl> <dbl> <dbl>
## 1 -Inf 0 1
## 2 2.22e-16 0 1
## 3 5.62e- 8 0 1.00
## 4 3.42e- 7 0 1.00
## 5 3.61e- 7 0 1.00
## 6 4.70e- 7 0 1.00
## 7 6.13e- 7 0 1.00
## 8 6.15e- 7 0 1.00
## 9 6.52e- 7 0 1.00
## 10 8.00e- 7 0 1.00
## # … with 22,700 more rows# Plot ROC curve
threshold_df %>%
autoplot()
# Calculate ROC AUC
roc_auc(nasa_results,
truth = hazardous,
.pred_True)## # A tibble: 1 × 3
## .metric .estimator .estimate
## <chr> <chr> <dbl>
## 1 roc_auc binary 0.123Random Forest
spec <- rand_forest(mtry = 4,
trees = 100,
min_n = 10) %>%
set_mode("classification") %>%
set_engine("ranger")Training the Forest
spec2 <- spec %>%
fit(hazardous ~ est_diameter_min + est_diameter_max + relative_velocity + miss_distance + absolute_magnitude, data = nasa_training)
#library("ranger")
#model_rf <- ranger(hazardous ~ ., data = nasa_training, probability = TRUE)
##data = nasa_training[complete.cases(nasa_training),]
#model_rfVariable Importance
We observe that the most important variables are “miss_distance” and “relative_velocity”, which are followed by “est_diameter_max”, “absolute_magnitude” and “est_diameter_min”.
rand_forest(mode = "classification") %>%
set_engine("ranger", importance = "impurity") %>%
fit(hazardous ~ ., data = nasa_training) %>%
vip::vip()
Predictions based on Random Forest Model
forest_predictions <- predict(spec2, new_data = nasa_test)
forest_predictions## # A tibble: 22,709 × 1
## .pred_class
## <fct>
## 1 False
## 2 False
## 3 False
## 4 False
## 5 False
## 6 False
## 7 False
## 8 False
## 9 False
## 10 False
## # … with 22,699 more rows#Combining predictions and truth value
pred_combined <- forest_predictions %>%
mutate(true_class = nasa_test$hazardous)
head(pred_combined)## # A tibble: 6 × 2
## .pred_class true_class
## <fct> <fct>
## 1 False False
## 2 False False
## 3 False True
## 4 False False
## 5 False False
## 6 False False#Confusion matrix
conf_mat(data = pred_combined, estimate = .pred_class, truth = true_class)## Truth
## Prediction False True
## False 20058 1475
## True 425 751Accuracy of Random Forest
# Calculate the accuracy
accuracy(pred_combined, truth = true_class,
estimate = .pred_class)## # A tibble: 1 × 3
## .metric .estimator .estimate
## <chr> <chr> <dbl>
## 1 accuracy binary 0.916Sensitivity of Random Forest
# Calculate the sensitivity
sens(pred_combined, truth = true_class,
estimate = .pred_class)## # A tibble: 1 × 3
## .metric .estimator .estimate
## <chr> <chr> <dbl>
## 1 sens binary 0.979Specificity of Random Forest
# Calculate the specificity
spec(pred_combined, truth = true_class,
estimate = .pred_class)## # A tibble: 1 × 3
## .metric .estimator .estimate
## <chr> <chr> <dbl>
## 1 spec binary 0.337Boosting the Random Forest
library("xgboost")
boost_spec <- boost_tree() %>%
set_mode("classification") %>%
set_engine("xgboost")
boost_spec## Boosted Tree Model Specification (classification)
##
## Computational engine: xgboostboost_model <- fit(boost_spec, formula = hazardous ~ est_diameter_min + est_diameter_max + relative_velocity + miss_distance + absolute_magnitude, data = nasa_training)
boost_model## parsnip model object
##
## ##### xgb.Booster
## raw: 44.9 Kb
## call:
## xgboost::xgb.train(params = list(eta = 0.3, max_depth = 6, gamma = 0,
## colsample_bytree = 1, colsample_bynode = 1, min_child_weight = 1,
## subsample = 1, objective = "binary:logistic"), data = x$data,
## nrounds = 15, watchlist = x$watchlist, verbose = 0, nthread = 1)
## params (as set within xgb.train):
## eta = "0.3", max_depth = "6", gamma = "0", colsample_bytree = "1", colsample_bynode = "1", min_child_weight = "1", subsample = "1", objective = "binary:logistic", nthread = "1", validate_parameters = "TRUE"
## xgb.attributes:
## niter
## callbacks:
## cb.evaluation.log()
## # of features: 5
## niter: 15
## nfeatures : 5
## evaluation_log:
## iter training_logloss
## 1 0.5000166
## 2 0.3939080
## ---
## 14 0.1814011
## 15 0.1795252set.seed(99)
folds <- vfold_cv(nasa_training, v = 3)
cv_results <- fit_resamples(boost_spec,
hazardous ~ est_diameter_min + est_diameter_max + relative_velocity + miss_distance + absolute_magnitude,
resamples = folds,
metrics = metric_set(yardstick::roc_auc, accuracy, sens, specificity))
collect_metrics(cv_results)## # A tibble: 4 × 6
## .metric .estimator mean n std_err .config
## <chr> <chr> <dbl> <int> <dbl> <chr>
## 1 accuracy binary 0.913 3 0.00111 Preprocessor1_Model1
## 2 roc_auc binary 0.915 3 0.000505 Preprocessor1_Model1
## 3 sens binary 0.996 3 0.000758 Preprocessor1_Model1
## 4 specificity binary 0.143 3 0.00724 Preprocessor1_Model1set.seed(100)
predictions <- boost_tree() %>%
set_mode("classification") %>%
set_engine("xgboost") %>%
fit(hazardous ~ ., data = nasa_training) %>%
predict(new_data = nasa_training, type = "prob") %>%
bind_cols(nasa_training)
predictions## # A tibble: 68,127 × 8
## .pred_False .pred_True est_diameter_min est_diameter_max relative_velocity
## <dbl> <dbl> <dbl> <dbl> <dbl>
## 1 0.980 0.0203 1198. 2679. 13569.
## 2 0.723 0.277 266. 594. 73589.
## 3 0.995 0.00501 36.4 81.3 34298.
## 4 0.706 0.294 172. 384. 27529.
## 5 0.784 0.216 350. 784. 56625.
## 6 0.713 0.287 253. 565. 58431.
## 7 0.693 0.307 153. 342. 64394.
## 8 0.995 0.00501 69.9 156. 38019.
## 9 0.977 0.0227 290. 649. 10402.
## 10 0.995 0.00501 44.1 98.6 70771.
## # … with 68,117 more rows, and 3 more variables: miss_distance <dbl>,
## # absolute_magnitude <dbl>, hazardous <fct># Calculate AUC
roc_auc(predictions,
truth = hazardous,
estimate = .pred_True)## # A tibble: 1 × 3
## .metric .estimator .estimate
## <chr> <chr> <dbl>
## 1 roc_auc binary 0.0771boost_spec <- boost_tree(
trees = 100,
learn_rate = tune(),
tree_depth = tune(),
sample_size = tune()) %>%
set_mode("classification") %>%
set_engine("xgboost")
# Create the tuning grid
tunegrid_boost <- grid_regular(parameters(boost_spec),
levels = 3)## Warning: `parameters.model_spec()` was deprecated in tune 0.1.6.9003.
## Please use `hardhat::extract_parameter_set_dials()` instead.tunegrid_boost## # A tibble: 27 × 3
## tree_depth learn_rate sample_size
## <int> <dbl> <dbl>
## 1 1 0.001 0.1
## 2 8 0.001 0.1
## 3 15 0.001 0.1
## 4 1 0.0178 0.1
## 5 8 0.0178 0.1
## 6 15 0.0178 0.1
## 7 1 0.316 0.1
## 8 8 0.316 0.1
## 9 15 0.316 0.1
## 10 1 0.001 0.55
## # … with 17 more rows# Create CV folds of training data
folds <- vfold_cv(nasa_training, v = 3)
# Tune along the grid
tune_results <- tune_grid(boost_spec,
hazardous ~ est_diameter_min + est_diameter_max + relative_velocity + miss_distance + absolute_magnitude,
resamples = folds,
grid = tunegrid_boost,
metrics = metric_set(yardstick::roc_auc))
# Plot the results
autoplot(tune_results)