Code for calculating CVRMSE of each case

Author: ClaireGyn

models/cal_cvrmse.py

@@ -0,0 +1,46 @@
+import numpy as np
+import pandas as pd
+
+
+# Function for calculating CVRMSE
+def cal_cvrmse(y_pred, y_true):
+	n_pred = np.shape(y_pred)[0]
+	cvrmse = 100*np.sqrt(np.sum(np.square(y_pred-y_true))/(n_pred-1))/np.mean(y_true)
+
+	return cvrmse 
+
+# Function for computing the mean of each column
+def col_means(y_pred_raw):
+	df = y_pred_raw
+	col_mean_df = df.mean(axis=0)
+
+	return col_mean_df
+
+
+name_lst = [case_name list]  # Put case names here
+case_num = len(name_lst)
+cvrmse_df = pd.DataFrame(columns=['CVRMSE'])  # Create a dataframe for storing CVRMSE results
+parent_fd = "Pata of results"  # Put the path of results here
+
+# Loop through each output of each case
+for i, name in enumerate(name_lst):
+    for j in range(case_num):
+        conf = yaml.load(open(os.path.join(parent_fd, "conf_{}_{}.yaml".format(name_lst[i], j+1))), Loader=yaml.FullLoader)
+        n_y = len(conf['vc_keys'])+len(conf['yc_keys'])
+        
+        for k in range(n_y):
+            # Read the pred data of each output of each case
+            y_pred_raw = pd.read_csv(os.path.join(os.getcwd(), '_case_{}_{}_y_pred{}.csv'.format(name_lst[i], j+1, k+1)), index_col=0)  
+            y_pred_raw = y_pred_raw.T.reset_index(drop=True)
+
+            # Read the DataField of each output of each case
+            y_true = pd.read_csv(os.path.join(os.getcwd(), 'Path_of_Datafield + {}_{}'.format(name_lst[i], j+1)))
+
+            # Calculate CVRMSE of each output of each case
+            cvrmse = cal_cvrmse(col_means(y_pred_raw.T), y_true.iloc[:, k])
+            cvrmse_df.loc['{}_{}_y_pred{}'.format(name_lst[i], j+1, k+1)] = cvrmse
+
+# Create folder for storing the results
+if not os.path.exists('./data'):
+    os.makedirs('./data')
+cvrmse_df.to_csv('./data/cvrmse_res.csv')