Business Context
A healthcare provider needs to run a diagnostic prediction model on a batch of new patient records. The dataset is known to be 'dirty': some records have missing test results (input variables), while others are for returning patients who already have a manually entered diagnosis (target variable). The goal is to score only the patients without a diagnosis, while gracefully handling those with incomplete data.
About the Set : neuralNet
Training of classical artificial neural networks.
Discover all actions of neuralNet Data Preparation
Creates a `patient_records` table. Patient P01 has a pre-existing diagnosis. P02 is a new patient with complete data. P03 is a new patient with a missing `BloodPressure` value. The model `diagnostic_model` is assumed to exist.
Copied!
| 1 | DATA mycas.patient_records(promote=yes); |
| 2 | LENGTH Diagnosis $20.; |
| 3 | INPUT PatientID Age BloodPressure Cholesterol Diagnosis $; |
| 4 | DATALINES; |
| 5 | 1 65 140 200 Type2Diabetes |
| 6 | 2 45 120 190 . |
| 7 | 3 55 . 220 . |
| 8 | 4 70 160 . . |
| 9 | ; |
| 10 | RUN; |
Étapes de réalisation
1
Run `annScore` with `impute=true`. This should prevent re-scoring patients who already have a value in the target variable `Diagnosis`.
Copied!
| 1 | PROC CAS; |
| 2 | neuralNet.annScore / |
| 3 | TABLE={name='patient_records'}, |
| 4 | modelTable={name='diagnostic_model'}, |
| 5 | casOut={name='diagnostic_results', replace=true}, |
| 6 | copyVars={'PatientID'}, |
| 7 | impute=true; |
| 8 | RUN; |
| 9 | QUIT; |
2
Fetch and examine the results to confirm the imputation logic. Patient P01's prediction should match their original diagnosis. Patients P02 and P03 should have new predictions, even though P03 had a missing input value.
Copied!
| 1 | PROC CAS; |
| 2 | TABLE.fetch / TABLE='diagnostic_results'; |
| 3 | RUN; |
| 4 | QUIT; |
Expected Result
The output table `mycas.diagnostic_results` will be created. For PatientID=1, the predicted value column `_NN_PredName_` should be 'Type2Diabetes', matching the source data. For PatientID=2 and PatientID=3, the `_NN_PredName_` column should contain a model-generated prediction. The action should complete without errors, demonstrating that the model's internal imputation handled the missing `BloodPressure` for P03 and missing `Cholesterol` for P04.