Patient Wise DAP

dap_prediction_LR.py

@@ -25,29 +25,30 @@ from dap.models.featuresmodel import DAPSVMClassifier, DAPRandomForestClassifier
 # from dap.models.sklearnmodel import SklearnModel
 from sklearn import svm
 
-from dap.crossval import kfold_split
+from dap.crossval import kfold_split, patient_wise_cv_fold
 from dap.ranking import randomforest_ranking, kbest_ranking
 # from dap_devel.dap.dap import DAP
 
 
 # %%
 SEED = 1234
-LAB_COL = 'Locoregional'
-FEATURESET = 'radiomics_features_CT_F_SVC_Locoregional' #'features_LR' #radiomics_features_F_SVC_Locoregional
+LAB_COL = 'label'
+PATIENT_COL = 'patient'
+FEATURESET = 'features_noTx_AUG' #'features_LR' #radiomics_features_F_SVC_Locoregional
 
 #%%
-LABELS_FILE = f'{PATH}/data/clinical_data.csv'
+LABELS_FILE = f'{PATH}/data/clinical_data_noTx.csv'
 NETWORK = 'LR_noTx_branch_wise_free_aug_CT_20191027-124913'
 
-FEATUREDIR = f'{PATH}/data/' #f'{PATH}/experiments/{NETWORK}/features' # # ## 
-OUT_DIR = f'{PATH}/predictions/{FEATURESET}'
+FEATUREDIR =  f'{PATH}/experiments/{NETWORK}/features' #f'{PATH}/data/' # ## 
+#OUT_DIR = f'{PATH}/predictions/{FEATURESET}'
+OUT_DIR = f'{PATH}/experiments/{NETWORK}'
 
 FEATURE_FILE_DEEP = f'{FEATUREDIR}/{FEATURESET}.csv'
-FEATURE_FILE_RADIOMICS = f'{FEATUREDIR}/{FEATURESET}.csv'
 
 
 # %%
-dataset = FeatureLabelCSVDataset(FEATURE_FILE_RADIOMICS, LABELS_FILE, LAB_COL, problem='binary', name=FEATURESET)
+dataset = CSVDataset(FEATURE_FILE_DEEP, LAB_COL, PATIENT_COL, problem='binary', name=FEATURESET)
 
 #%%
 metrics = BinaryMetrics(reference='MCC')
@@ -56,8 +57,9 @@ dap_settings = {'stratified_test': True,
                 'ratio_test': 0.2,
                 'stratified_valid': True,
                 'ratio_valid': 0.2,
-                'fold_method': kfold_split,
-                'cv_n': 10}
+                'fold_method': patient_wise_cv_fold,
+                'cv_n': 10,
+                'patient_wise_split': True}
 
 # %% [markdown]
 # ## LSVM
@@ -69,7 +71,7 @@ model = DAPSVMClassifier({'rank_features': False, 'scale_features': True, 'kerne
 # %%
 DAP = dap.DAP(dap_settings, dataset, model, metrics, name='kfold', seed=SEED)
 
-n_feat = [0.001, 0.002, 0.005, 0.01, 0.02, 0.05, 0.1, 0.2, 0.5, 1]
+n_feat = [0.25, 0.5, 0.75, 1]
 parameters = {'C':[0.001, 0.01, 0.1, 1, 10, 100, 1000], 'n_feat': n_feat} # [0.001, 0.01, 0.1, 1, 10, 100, 1000]
 param_grid = dap.dap.create_param_grid(parameters)
 
@@ -84,24 +86,24 @@ DAP.save(OUT_DIR)
 # ## Random Forest
 
 # %%
-model = DAPRandomForestClassifier({'rank_features': False, 'scale_features': True, 'n_estimators': 500, 'random_labels': False}, name='DAPRFClassifier')
+# model = DAPRandomForestClassifier({'rank_features': False, 'scale_features': True, 'n_estimators': 500, 'random_labels': False}, name='DAPRFClassifier2')
 
 
 # %%
-DAP = dap.DAP(dap_settings, dataset, model, metrics, name='kfold')
+# DAP = dap.DAP(dap_settings, dataset, model, metrics, name='kfold2')
 
 
 #%%
-n_feat = [0.001, 0.002, 0.005, 0.01, 0.02, 0.05, 0.1, 0.2, 0.5, 1]
-parameters = {'n_feat': n_feat} # [0.001, 0.01, 0.1, 1, 10, 100, 1000]
-param_grid = dap.dap.create_param_grid(parameters)
+# n_feat = [0.001, 0.002, 0.005, 0.01, 0.02, 0.05, 0.1, 0.2, 0.5, 1]
+# parameters = {'n_feat': n_feat} # [0.001, 0.01, 0.1, 1, 10, 100, 1000]
+# param_grid = dap.dap.create_param_grid(parameters)
 
-DAP.fit(param_grid)
+# DAP.fit(param_grid)
 
 
 # %%
-DAP.predict_on_test()
-DAP.save(OUT_DIR)
+# DAP.predict_on_test()
+# DAP.save(OUT_DIR)
 
 
 # %%