Non stratified experiments

Remove reference column in analysis

analysis_mysql.py

@@ -15,7 +15,7 @@ models_tree = [
 ]
 models_ensemble = ["odte", "adaBoost", "bagging", "TBRaF", "TBRoF", "TBRRoF"]
 title = "Best model results"
-lengths = (30, 9, 11, 11, 11, 11, 11, 11)
+lengths = (30, 12, 12, 12, 12, 12, 12)
 
 
 def parse_arguments() -> Tuple[str, str, str, bool, bool]:
@@ -63,7 +63,7 @@ def report_header_content(title, experiment, model_type):
     output += "*" * length + "\n\n"
     lines = ""
     for item, data in enumerate(fields):
-        output += f"{fields[item]:{lengths[item]}} "
+        output += f"{fields[item]:^{lengths[item]}} "
         lines += "=" * lengths[item] + " "
     output += f"\n{lines}"
     return output
@@ -80,31 +80,17 @@ def report_header(title, experiment, model_type):
 def report_line(line):
     output = f"{line['dataset']:{lengths[0] + 5}s} "
     data = models.copy()
-    data.insert(0, "reference")
     for key, model in enumerate(data):
         output += f"{line[model]:{lengths[key + 1]}s} "
     return output
 
 
 def report_footer(agg):
-    print(
-        TextColor.GREEN
-        + f"we have better results {agg['better']['items']:2d} times"
-    )
-    print(
-        TextColor.RED
-        + f"we have worse  results {agg['worse']['items']:2d} times"
-    )
+    length = sum(lengths) + len(lengths) - 1
+    print("-" * length)
     color = TextColor.LINE1
     for item in models:
-        print(
-            color + f"{item:10s} used {agg[item]['items']:2d} times ", end=""
-        )
-        print(
-            color + f"better than reference {agg[item]['better']:2d} times ",
-            end="",
-        )
-        print(color + f"worse {agg[item]['worse']:2d} times ", end="")
+        print(color + f"{item:10s} ", end="")
         print(color + f"best of models {agg[item]['best']:2d} times")
         color = (
             TextColor.LINE2 if color == TextColor.LINE1 else TextColor.LINE1
@@ -115,7 +101,7 @@ def report_footer(agg):
 dbh = MySQL()
 database = dbh.get_connection()
 dt = Datasets(False, False, "tanveer")
-fields = ("Dataset", "Reference")
+fields = ("Dataset",)
 models = models_tree if model_type == "tree" else models_ensemble
 for item in models:
     fields += (f"{item}",)
@@ -127,9 +113,6 @@ for item in [
     "worse",
 ] + models:
     agg[item] = {}
-    agg[item]["items"] = 0
-    agg[item]["better"] = 0
-    agg[item]["worse"] = 0
     agg[item]["best"] = 0
 if csv_output:
     f = open(report_csv, "w")
@@ -143,22 +126,13 @@ for dataset in dt:
     for model in models:
         record = dbh.find_best(dataset[0], model, experiment)
         if record is None:
-            line[model] = color + "-" * 9 + "  "
+            line[model] = color + "-" * 12
         else:
             reference = record[13]
             accuracy = record[5]
+            acc_std = record[11]
             find_one = True
-            agg[model]["items"] += 1
-            if accuracy > reference:
-                sign = "+"
-                agg["better"]["items"] += 1
-                agg[model]["better"] += 1
-            else:
-                sign = "-"
-                agg["worse"]["items"] += 1
-                agg[model]["worse"] += 1
-            item = f"{accuracy:9.7} {sign}"
-            line["reference"] = f"{reference:9.7}"
+            item = f"{accuracy:.4f}±{acc_std:.3f}"
             if accuracy == max_accuracy:
                 line[model] = (
                     TextColor.GREEN + TextColor.BOLD + item + TextColor.ENDC