Skip to main content

6  Data checks and validation

Data inspection establishes the interpretation of each column, the unit of observation, and the outcome. A data dictionary supplies units, category meanings, missing-value codes, and the relationship between repeated cases.

6.1 Inspect a dataset

check_data(iris)
  iris: A data.frame with 150 rows and 5 columns.

  Data types
  * 4 numeric features
  * 0 integer features
  * 1 factor, which is not ordered
  * 0 character features
  * 0 date features

  Issues
  * 0 constant features
  * 1 duplicate case
  * 0 missing values

  Recommendations
  * Consider removing the duplicate case.

check_data() summarizes types, missingness, distinct values, and duplicate rows. An identical row can represent a duplicate record or a distinct case with identical measurements; its treatment depends on the study design.

6.2 Create a data profile

profile <- data_profile(iris, fingerprint = TRUE)
profile@columns
          name       dtype n_distinct n_missing
1 Sepal.Length      number         35         0
2  Sepal.Width      number         23         0
3 Petal.Length      number         43         0
4  Petal.Width      number         22         0
5      Species categorical          3         0
profile@level_counts
   column      level  n
1 Species     setosa 50
2 Species versicolor 50
3 Species  virginica 50

data_profile() returns a typed description of dataset dimensions, column types, missingness, distinct values, and factor levels. Optional fingerprints identify the data represented by the profile.

6.3 Check a configuration against data

config <- setup_SuperConfig(
  outcome = "Species",
  hyperparameters = setup_CART(),
  execution_config = setup_SerialExecution(seed = 2026L)
)
findings <- validate_config(config, data = iris)
findings
<Diagnostics>
  No problems found.

validate_config() returns structured diagnostics, with severity and corrective information where available. Configuration validity and data compatibility are separate checks. A recipe without bound data can be valid while data-dependent checks remain pending.

train(preflight = TRUE) runs the configuration and data checks before fitting. Error findings stop training; warning findings are reported.

mod <- train(
  iris,
  hyperparameters = setup_CART(),
  execution_config = setup_SerialExecution(seed = 2026L),
  preflight = TRUE
)
2026-10-08 02:53:08 Checking data is ready for training... ✔ [check_supervised]
2026-10-08 02:53:08 ▶ [train]
2026-10-08 02:53:08 Training set: 150 cases x 4 features. [summarize_supervised]
2026-10-08 02:53:08 // CPU | serial | 1 worker: algorithm 1 thread, tuning 1, outer resampling 1 [msg_resources]
2026-10-08 02:53:08 Shared memory (auto): not sharing training data (7.3 kB) -- this run transfers nothing to workers. [report_shared_memory]
2026-10-08 02:53:08 Training CART Classification... [train]
2026-10-08 02:53:09 Checking data is ready for training... ✔ [check_supervised]

<Classification>
CART (CART decision tree (rpart))

  <Training Classification Metrics>
                     Predicted
          Reference  setosa  versicolor  virginica  
             setosa      50           0          0
         versicolor       0          47          3
          virginica       0           1         49

                     Overall  
  Balanced Accuracy  0.973  
                 F1  0.973  
           Accuracy  0.973  
                     Setosa  Versicolor  Virginica  
        Sensitivity  1.000   0.940       0.980    
        Specificity  1.000   0.990       0.970    
  Balanced Accuracy  1.000   0.965       0.975    
                PPV  1.000   0.979       0.942    
                NPV  1.000   0.971       0.990    
                 F1  1.000   0.959       0.961    

2026-10-08 02:53:09 Done in 0.79 seconds. [train]
© 2026 E.D. Gennatas