add visualization for classification and clustering

Author: kinkunchan

.DS_Store

@@ -0,0 +1,101 @@
+from ..exceptions import *
+import networkx as nx
+import pandas as pd
+import matplotlib.pyplot as plt
+
+def visualize_classification(list_results, metric="f1_macro", error_bar=True, rotation_deg=45):
+    """
+    Visualize the result for classification using bar plot.
+
+    Arguments:
+    list_results {list of tuples} --  [("method_name", dict_result)], where the dict_result is the returned dict
+                                      from the perform_classification() function
+    metric -- The metric to use for making the plot. Please choose from ["accuracy", "f1_macro", "f1_micro", "f1_macro", "auc_micro", "auc_macro"]
+    error_bar -- bool. Whether to include the error bar in the plot.
+    rotation_deg -- int. The rotation degree for the xtick labels
+    """
+
+    # Check that metric belongs to the classification results
+    if metric not in ["accuracy", "f1_macro", "f1_micro", "f1_macro", "auc_micro", "auc_macro"]:
+        raise MethodKeywordUnAllowedException("Please choose metric from [accuracy, f1_macro, f1_micro, f1_macro, auc_micro, auc_macro].")
+
+    if len(list_results) == 0:
+        raise InputFormatErrorException("Input length 0!")
+
+    for cur_item in list_results:
+        if (len(cur_item) != 2):
+            raise InputFormatErrorException("Please input the results as list of tuples, i.e. [(\"method_name\", dict_result)]")
+
+        if (not isinstance(cur_item[0], str)) or (not isinstance(cur_item[1], dict)):
+            raise InputFormatErrorException("Please input the results as list of tuples, i.e. [(\"method_name\", dict_result)]")
+
+        if "overall" not in cur_item[1]:
+            raise InputFormatErrorException("Invalid input. Please make sure that the input result is generated from perform_classification() or perform_clustering()")
+
+    
+    list_methods = [i[0] for i in list_results]
+    list_evaluation = [i[1]['overall'][metric]['mean'] for i in list_results]
+    list_error = [i[1]['overall'][metric]['std'] for i in list_results]
+    plt.close()
+    plt.figure(figsize=(8, 2.5), dpi=300)
+    plt.style.use('ggplot')
+    if error_bar:
+        plt.bar(list_methods,
+                list_evaluation,
+                yerr = list_error,
+                capsize=10)
+    else:
+        plt.bar(list_methods,
+                list_evaluation,
+                capsize=10)
+    plt.ylim(0, 1)
+    plt.ylabel(metric)
+    plt.xlabel("Methods")
+    plt.xticks(rotation=rotation_deg)
+    plt.show()
+
+
+def visualize_clustering(list_results, metric="purity", rotation_deg=45):
+    """
+    Visualize the result for clustering using bar plot.
+
+    Arguments:
+    list_results {list of tuples} --  [("method_name", dict_result)], where the dict_result is the returned dict
+                                      from the perform_clustering() function
+    metric -- The metric to use for making the plot. Please choose from ["purity", "nmi"]
+    rotation_deg -- int. The rotation degree for the xtick labels
+    """
+
+    # Check that metric belongs to the classification results
+    if metric not in ["purity", "nmi"]:
+        raise MethodKeywordUnAllowedException("Please choose metric from [purity, nmi].")
+
+    if len(list_results) == 0:
+        raise InputFormatErrorException("Input length 0!")
+
+    for cur_item in list_results:
+        if (len(cur_item) != 2):
+            raise InputFormatErrorException("Please input the results as list of tuples, i.e. [(\"method_name\", dict_result)]")
+
+        if (not isinstance(cur_item[0], str)) or (not isinstance(cur_item[1], dict)):
+            raise InputFormatErrorException("Please input the results as list of tuples, i.e. [(\"method_name\", dict_result)]")
+
+        if "overall" not in cur_item[1]:
+            raise InputFormatErrorException("Invalid input. Please make sure that the input result is generated from perform_classification() or perform_clustering()")
+
+    
+    list_methods = [i[0] for i in list_results]
+    list_evaluation = [i[1]['overall'][metric] for i in list_results]
+    plt.close()
+    plt.figure(figsize=(8, 2.5), dpi=300)
+    plt.style.use('ggplot')
+    
+    plt.bar(list_methods,
+            list_evaluation,
+            capsize=10)
+    plt.ylim(0, 1)
+    plt.ylabel(metric)
+    plt.xlabel("Methods")
+    plt.xticks(rotation=rotation_deg)
+    plt.show()
+