fixing stanislavs and lennarts implementation so that they always give the same result.

code/functions.py

@@ -195,8 +195,15 @@ def ocel_get_object_distances_lh(ocel, object_type, weights_per_attribute) -> di
     # NORMALIZING / WEIGHTING / AVERAGING
     df_distance_matrix = df_normalize_columns(df_distance_matrix, list(attr_func_map.keys()))
     df_distance_matrix = df_weight_columns(df_distance_matrix, weights_per_attribute)
-    df_distance_matrix['distance_avg'] = df_distance_matrix[list(attr_func_map.keys())].mean(axis=1)
-    # df_distance_matrix = df_normalize_columns(df_distance_matrix, ['distance_avg']) -> leads to wrong results compared to stanislav
+    
+    divisor = 0.0
+    for attribute in list(attr_func_map.keys()):
+        divisor += weights_per_attribute.get(attribute, 1.0)
+    if divisor == 0.0: divisor = 1.0
+    df_distance_matrix['distance_avg'] = df_distance_matrix[list(attr_func_map.keys())].sum(axis=1) / divisor
+    # df_distance_matrix['distance_avg'] = df_distance_matrix[list(attr_func_map.keys())].mean(axis=1) -> mean over 3 values is not the same as weighting and dividing by the weights sum
+
+    # df_distance_matrix = df_normalize_columns(df_distance_matrix, ['distance_avg']) -> leads to wrong results compared to stanislav -> but maybe useful normalization
     df_distance_matrix = df_distance_matrix.drop(list(attr_func_map.keys()), axis=1) # single distances not relevant anymore
     
     # Creating the matrix-index->ocel:oid map
@@ -335,47 +342,32 @@ def distance_matrix(df, weights):
         single_column_df = df[columns[i]]
         matrix = matrix + calc_distance(single_column_df, i, n, weights)
         
+    print('weight-sum:' + str(sum(weights)))
     return matrix / sum(weights)
 
 
 
 # Helper function for the distance matrix calculation
 def calc_distance(df, weight_index, n, weights):
-    matrix = np.empty([n, n], dtype=float)
+    matrix = np.zeros([n, n], dtype=float)
     npm = df.values
     
+    distance_func = lambda x,y: abs(x-y)
     if (isinstance(npm[0], str) and df.name != "cflow"):
-        for j in range(n):
-            matrix[j][j] = 0
-            first_elem = npm[j]
-            for k in range(j+1,n):
-                matrix[j][k] = first_elem == npm[k]
-                matrix[k][j] = matrix[j][k]
-                
+        distance_func = lambda x,y: x!=y
     elif (isinstance(npm[0], float)):
-        matrix_temp = np.empty([n], dtype=float)
-        for j in range(n):
-            first_elem = npm[j]
-            for k in range(j + 1, n):
-                matrix_temp[k] = abs(first_elem - npm[k])
-            for m in range(j):
-                 matrix_temp[m] = matrix[m][j]
-            matrix_temp[j] = 0
-            matrix[j] = matrix_temp
-        matrix = matrix / matrix.max()
-        
+        distance_func = lambda x,y: abs(x-y)
     else:
-        matrix_temp = np.empty([n], dtype=float)
-        for j in range(n):
+        distance_func = lev
+    
+    for j in range(n):
             first_elem = npm[j]
-            for k in range(j + 1, n):
-                matrix_temp[k] = lev(first_elem, npm[k])
-            for m in range(j):
-                 matrix_temp[m] = matrix[m][j]
-            matrix_temp[j] = 0
-            matrix[j] = matrix_temp
-        matrix = matrix / matrix.max()
-        
+            for k in range(j+1,n):
+                matrix[j, k] = distance_func(first_elem, npm[k])
+    matrix = matrix + matrix.T
+    matrix = matrix / matrix.max()
+
+    print('current weight: ' + str(weights[weight_index]))
     return matrix * weights[weight_index]
 
 
@@ -416,6 +408,7 @@ def ocel_get_object_distances_sy(ocel, object_type, weights_per_attribute) -> di
     # Drop oid because it is not needed for the matrix calculation
     oid_filtered.drop(['ocel:oid'], axis = 1, inplace = True)
 
+    # weight-dictionary to array
     columns = list(oid_filtered.columns)
     weights = []
     for column in columns:

code/main.py

@@ -14,8 +14,9 @@ from sklearn import cluster as skcluster
 # PARAMETERS
 p_ocel_file         = './data/ocel.jsonocel'
 p_mode              = 'existence' # all | existence
-p_object_type       = 'customers' # object-type-name, options depend on the data
+p_object_type       = 'products' # object-type-name, options depend on the data
 p_attr_weights = {
+    'producer': 2
 } # attributes that are not given in the data are not used
 p_clustering_mode   = 'kmeans' # optional, default: kmeans
 p_cluster_count     = np.NaN # cluster-count (optional, default: np.NaN which leads to automatic k determination)
@@ -52,8 +53,10 @@ assert p_clustering_mode in algorithms.keys(), 'The given clustering mode "' + p
 
 ocel = pm4py.read_ocel(p_ocel_file)
 
-# res_lh = f.ocel_get_object_distances_lh(ocel, p_object_type, p_attr_weights)
 res = f.ocel_get_object_distances_sy(ocel, p_object_type, p_attr_weights)
+# res_lh = f.ocel_get_object_distances_lh(ocel, p_object_type, p_attr_weights)
+
+# print(res['distances']-res_lh['distances'])
 
 duration = datetime.datetime.now() - start_ts
 print('-------------------------------------------------------')