Breast Cancer Wisconsin - Random Forest Classification Report
Data Science Report
August 04, 2026
1. Data Ingestion
bc_data <- read.csv("breast_cancer.csv", stringsAsFactors = FALSE)
bc_data$id <- NULL
empty_cols <- names(bc_data)[names(bc_data) == "" | grepl("^X$|^X\\.", names(bc_data))]
bc_data[empty_cols] <- NULL
bc_data <- bc_data[, colSums(is.na(bc_data)) < nrow(bc_data)]
numeric_cols <- sapply(bc_data, is.numeric)
for (col in names(bc_data)[numeric_cols]) {
if (any(is.na(bc_data[[col]]))) {
bc_data[[col]][is.na(bc_data[[col]])] <- median(bc_data[[col]], na.rm = TRUE)
}
}
bc_data$diagnosis <- factor(bc_data$diagnosis, levels = c("B", "M"),
labels = c("Benign", "Malignant"))
bc_data <- bc_data[!is.na(bc_data$diagnosis), ]kable(data.frame(Metric = c("Rows", "Columns"),
Value = c(nrow(bc_data), ncol(bc_data))),
caption = "Dataset Dimensions")| Metric | Value |
|---|---|
| Rows | 569 |
| Columns | 31 |
kable(as.data.frame(table(bc_data$diagnosis)),
col.names = c("Diagnosis", "Count"),
caption = "Class Distribution")| Diagnosis | Count |
|---|---|
| Benign | 357 |
| Malignant | 212 |
ggplot(bc_data, aes(x = diagnosis, fill = diagnosis)) +
geom_bar() +
geom_text(stat = "count", aes(label = after_stat(count)), vjust = -0.5, size = 4.5) +
scale_fill_manual(values = c("Benign" = "#4C9F70", "Malignant" = "#D9534F")) +
labs(title = "Class Distribution", x = "Diagnosis", y = "Number of Cases") +
theme_report +
theme(legend.position = "none")
2. Data Partitioning
set.seed(42)
train_index <- createDataPartition(bc_data$diagnosis, p = 0.80, list = FALSE)
train_data <- bc_data[train_index, ]
test_data <- bc_data[-train_index, ]
kable(data.frame(Set = c("Training", "Test"),
Rows = c(nrow(train_data), nrow(test_data))),
caption = "Train / Test Split (80/20)")| Set | Rows |
|---|---|
| Training | 456 |
| Test | 113 |
3. Model Fitting
set.seed(42)
tuned_mtry <- tuneRF(
x = train_data[, -which(names(train_data) == "diagnosis")],
y = train_data$diagnosis,
ntreeTry = 500,
stepFactor = 1.5,
improve = 0.01,
trace = FALSE,
plot = FALSE
)
best_mtry <- tuned_mtry[which.min(tuned_mtry[, "OOBError"]), "mtry"]Selected mtry: 3
rf_model <- randomForest(
diagnosis ~ .,
data = train_data,
ntree = 1000,
mtry = best_mtry,
importance = TRUE
)
saveRDS(rf_model, "rf_model.rds")
print(rf_model)##
## Call:
## randomForest(formula = diagnosis ~ ., data = train_data, ntree = 1000, mtry = best_mtry, importance = TRUE)
## Type of random forest: classification
## Number of trees: 1000
## No. of variables tried at each split: 3
##
## OOB estimate of error rate: 4.39%
## Confusion matrix:
## Benign Malignant class.error
## Benign 279 7 0.02447552
## Malignant 13 157 0.07647059The trained model is saved to rf_model.rds in the
working directory, and can be reloaded later without retraining:
rf_model <- readRDS("rf_model.rds")
predict(rf_model, newdata = new_patient_data)4. Model Evaluation
rf_predictions <- predict(rf_model, newdata = test_data)
conf_matrix <- confusionMatrix(
rf_predictions,
test_data$diagnosis,
positive = "Malignant"
)
saveRDS(conf_matrix, "confusion_matrix.rds")
conf_matrix## Confusion Matrix and Statistics
##
## Reference
## Prediction Benign Malignant
## Benign 69 0
## Malignant 2 42
##
## Accuracy : 0.9823
## 95% CI : (0.9375, 0.9978)
## No Information Rate : 0.6283
## P-Value [Acc > NIR] : <2e-16
##
## Kappa : 0.9625
##
## Mcnemar's Test P-Value : 0.4795
##
## Sensitivity : 1.0000
## Specificity : 0.9718
## Pos Pred Value : 0.9545
## Neg Pred Value : 1.0000
## Prevalence : 0.3717
## Detection Rate : 0.3717
## Detection Prevalence : 0.3894
## Balanced Accuracy : 0.9859
##
## 'Positive' Class : Malignant
## Confusion Matrix Heatmap
cm_table <- as.data.frame(conf_matrix$table)
colnames(cm_table) <- c("Predicted", "Actual", "Freq")
ggplot(cm_table, aes(x = Actual, y = Predicted, fill = Freq)) +
geom_tile(color = "white", linewidth = 1) +
geom_text(aes(label = Freq), color = "white", size = 8, fontface = "bold") +
scale_fill_gradient(low = "#8FBFDE", high = "#1F4E79") +
labs(title = "Confusion Matrix", x = "Actual Diagnosis", y = "Predicted Diagnosis") +
theme_report +
theme(legend.position = "none")
Key Performance Metrics
metrics_df <- data.frame(
Metric = c("Accuracy", "Sensitivity (Malignant Recall)", "Specificity"),
Value = round(c(conf_matrix$overall["Accuracy"],
conf_matrix$byClass["Sensitivity"],
conf_matrix$byClass["Specificity"]), 4)
)
kable(metrics_df, caption = "Test Set Performance")| Metric | Value | |
|---|---|---|
| Accuracy | Accuracy | 0.9823 |
| Sensitivity | Sensitivity (Malignant Recall) | 1.0000 |
| Specificity | Specificity | 0.9718 |
- Accuracy: overall proportion of correct predictions across both classes.
- Sensitivity: of all actual malignant cases, the proportion correctly flagged — the most clinically important number, since a missed malignant case is the costliest error.
- Specificity: of all actual benign cases, the proportion correctly cleared.
ggplot(metrics_df, aes(x = Metric, y = Value, fill = Metric)) +
geom_col(width = 0.6) +
geom_text(aes(label = scales::percent(Value, accuracy = 0.1)), vjust = -0.5, size = 4.5) +
scale_y_continuous(limits = c(0, 1.05), labels = scales::percent) +
scale_fill_manual(values = c("#1F4E79", "#D9534F", "#4C9F70")) +
labs(title = "Test Set Performance Metrics", x = NULL, y = "Score") +
theme_report +
theme(legend.position = "none",
axis.text.x = element_text(angle = 15, hjust = 1))
ROC Curve
rf_probs <- predict(rf_model, newdata = test_data, type = "prob")
roc_obj <- roc(response = test_data$diagnosis,
predictor = rf_probs[, "Malignant"],
levels = c("Benign", "Malignant"),
direction = "<")
auc_val <- round(auc(roc_obj), 4)
plot(roc_obj,
main = paste0("ROC Curve (AUC = ", auc_val, ")"),
col = "#1F4E79", lwd = 3,
legacy.axes = TRUE)
abline(a = 0, b = 1, lty = 2, col = "gray60")
The ROC curve shows the tradeoff between sensitivity and specificity across all possible classification thresholds. An AUC (area under the curve) close to 1.0 indicates the model separates benign from malignant cases very well; 0.5 would indicate no better than random guessing.
5. Feature Importance
importance_df <- as.data.frame(round(importance(rf_model), 3))
kable(importance_df, caption = "Variable Importance Scores")| Benign | Malignant | MeanDecreaseAccuracy | MeanDecreaseGini | |
|---|---|---|---|---|
| radius_mean | 12.200 | 9.580 | 14.810 | 9.426 |
| texture_mean | 9.540 | 11.339 | 13.903 | 3.591 |
| perimeter_mean | 11.886 | 9.362 | 14.389 | 10.630 |
| area_mean | 14.463 | 9.339 | 15.967 | 11.822 |
| smoothness_mean | 3.759 | 9.487 | 10.108 | 1.690 |
| compactness_mean | 7.622 | 6.804 | 10.185 | 3.818 |
| concavity_mean | 10.555 | 11.049 | 15.351 | 10.667 |
| concave.points_mean | 14.498 | 15.465 | 19.711 | 19.383 |
| symmetry_mean | 2.073 | 6.083 | 6.588 | 1.006 |
| fractal_dimension_mean | 6.412 | 2.553 | 6.740 | 1.412 |
| radius_se | 9.762 | 7.989 | 12.616 | 4.994 |
| texture_se | 4.495 | 4.218 | 6.182 | 1.309 |
| perimeter_se | 10.427 | 9.676 | 14.146 | 5.248 |
| area_se | 15.152 | 11.325 | 18.681 | 11.276 |
| smoothness_se | 2.807 | 0.065 | 2.344 | 1.257 |
| compactness_se | 7.110 | 2.090 | 7.350 | 1.823 |
| concavity_se | 6.739 | 5.378 | 8.638 | 2.270 |
| concave.points_se | 6.976 | 3.215 | 7.853 | 2.106 |
| symmetry_se | 3.506 | 2.038 | 4.160 | 1.089 |
| fractal_dimension_se | 3.061 | -0.787 | 2.010 | 1.417 |
| radius_worst | 17.810 | 15.278 | 21.863 | 18.631 |
| texture_worst | 11.425 | 13.369 | 16.371 | 3.767 |
| perimeter_worst | 17.038 | 15.039 | 21.162 | 19.923 |
| area_worst | 17.822 | 16.286 | 22.461 | 19.553 |
| smoothness_worst | 10.016 | 11.760 | 14.205 | 3.221 |
| compactness_worst | 8.632 | 8.718 | 12.425 | 6.053 |
| concavity_worst | 11.687 | 13.906 | 17.776 | 9.603 |
| concave.points_worst | 16.993 | 17.340 | 22.737 | 20.964 |
| symmetry_worst | 7.198 | 9.309 | 11.445 | 2.537 |
| fractal_dimension_worst | 5.848 | 4.532 | 7.340 | 2.103 |
varImpPlot(
rf_model,
main = "Random Forest - Feature Importance",
n.var = min(15, nrow(importance_df)),
pch = 19,
color = "steelblue"
)
6. Session Info
sessionInfo()## R version 4.5.2 (2025-10-31)
## Platform: aarch64-apple-darwin20
## Running under: macOS Ventura 13.3
##
## Matrix products: default
## BLAS: /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib
## LAPACK: /Library/Frameworks/R.framework/Versions/4.5-arm64/Resources/lib/libRlapack.dylib; LAPACK version 3.12.1
##
## locale:
## [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
##
## time zone: America/Chicago
## tzcode source: internal
##
## attached base packages:
## [1] stats graphics grDevices utils datasets methods base
##
## other attached packages:
## [1] pROC_1.19.0.1 knitr_1.51 caret_7.0-1
## [4] lattice_0.22-7 ggplot2_4.0.3 randomForest_4.7-1.2
##
## loaded via a namespace (and not attached):
## [1] gtable_0.3.6 xfun_0.60 bslib_0.12.0
## [4] recipes_1.3.3 vctrs_0.7.3 tools_4.5.2
## [7] generics_0.1.4 stats4_4.5.2 parallel_4.5.2
## [10] proxy_0.4-29 tibble_3.3.1 pkgconfig_2.0.3
## [13] ModelMetrics_1.2.2.2 Matrix_1.7-4 data.table_1.18.4
## [16] RColorBrewer_1.1-3 S7_0.2.2 lifecycle_1.0.5
## [19] compiler_4.5.2 farver_2.1.2 stringr_1.6.0
## [22] codetools_0.2-20 htmltools_0.5.9 class_7.3-23
## [25] sass_0.4.10 yaml_2.3.12 prodlim_2026.03.11
## [28] pillar_1.11.1 jquerylib_0.1.4 MASS_7.3-65
## [31] cachem_1.1.0 gower_1.0.2 iterators_1.0.14
## [34] rpart_4.1.24 foreach_1.5.2 nlme_3.1-168
## [37] parallelly_1.48.0 lava_1.9.2 tidyselect_1.2.1
## [40] digest_0.6.39 stringi_1.8.7 future_1.75.0
## [43] dplyr_1.2.1 reshape2_1.4.5 purrr_1.2.2
## [46] listenv_1.0.0 labeling_0.4.3 splines_4.5.2
## [49] fastmap_1.2.0 grid_4.5.2 cli_3.6.6
## [52] magrittr_2.0.5 survival_3.8-3 e1071_1.7-17
## [55] future.apply_1.20.2 withr_3.0.3 scales_1.4.0
## [58] lubridate_1.9.5 timechange_0.4.0 rmarkdown_2.31
## [61] globals_0.19.1 nnet_7.3-20 timeDate_4052.112
## [64] evaluate_1.0.5 hardhat_1.4.3 rlang_1.3.0
## [67] Rcpp_1.1.2 glue_1.8.1 ipred_0.9-15
## [70] rstudioapi_0.19.0 jsonlite_2.0.0 R6_2.6.1
## [73] plyr_1.8.9