Add interface to display confusion matrices

main.py

@@ -2,7 +2,7 @@
 from threading import Thread
 from src.fires_info import get_fires_dataframe
 from src.location_info import get_counties_geodf,get_fips_codes_dataframe
-from src.prediction import get_fips_encoder, get_fips_model, get_lonlat_model, joblib_objects_unpacked
+from src.prediction import get_fips_encoder, get_fips_model, get_lonlat_model, joblib_objects_present
 
 # Load the database data and pygris geographical data in the background
 def fires_info_thread_target():
@@ -20,7 +20,7 @@ def location_info_thread_target():
 fires_info_thread.start()
 location_info_thread.start()
 
-enable_ml = joblib_objects_unpacked()
+enable_ml = joblib_objects_present()
 if enable_ml:
   # Load the ML prediction models in the background
   def prediction_thread_target():

src/gui/predictions.py

@@ -4,6 +4,7 @@
 from .custom_widgets import NotebookFrame, Title, Subtitle, DefaultEntry, DatetimeEntry
 from ..location_info import get_fips_codes_dataframe, are_coordinates_inside_usa
 from ..prediction import run_fips_model_prediction, run_lonlat_model_prediction
+from ..bindings import plot_lonlat_confusion_matrix, plot_fipscode_confusion_matrix
 import datetime
 
 class PredictionsFrame(NotebookFrame):
@@ -47,9 +48,9 @@ def __init__(self, *args, **kwargs):
     self.set_up_state_county_variables()
 
     title = Title(self, text='Predictions')
-    subframe = ttk.LabelFrame(self, text='Predict cause of fire by giving some parameters')
+    subframe = ttk.LabelFrame(self, text='Predict cause of fire by giving some parameters', padding=5)
 
-    input_frame = ttk.Frame(subframe, padding=5)
+    input_frame = ttk.Frame(subframe)
     discovery_datetime_label = ttk.Label(input_frame, text='Discovery datetime: ')
     self.discovery_datetime_entry = DatetimeEntry(input_frame)
     contained_datetime_label = ttk.Label(input_frame, text='Contained datetime: ')
@@ -71,8 +72,7 @@ def __init__(self, *args, **kwargs):
     latitude_label = ttk.Label(location_input_frame, text='Latitude: ')
     self.latitude_entry = DefaultEntry(location_input_frame, default_text='42.124', validate='key', validatecommand=(self.register(PredictionsFrame.check_latitude), '%P'))
     predict_button = ttk.Button(input_frame, text='Predict fire cause', command=self.run_prediction)
-    display_frame = ttk.Frame(subframe, padding=5)
-    display_frame_subtitle = Subtitle(display_frame, text='Predicted Category Results')
+    display_frame = ttk.LabelFrame(subframe, padding=5, text='Predicted Category Results')
     predicted_category_label = ttk.Label(display_frame, text='Top predicted category:')
     self.predicted_category_var = StringVar()
     predicted_category = ttk.Label(display_frame, textvariable=self.predicted_category_var, font=('Helvetica', 12))
@@ -99,11 +99,14 @@ def __init__(self, *args, **kwargs):
     self.probability_3_percent_var = StringVar()
     probability_3_percent = ttk.Label(probabilities_frame, textvariable=self.probability_3_percent_var)
     probability_3_percent_label = ttk.Label(probabilities_frame, text='%')
+    confusion_matrix_buttons_frame = ttk.LabelFrame(subframe, text="ML Models' Confusion Matrices")
+    fips_confusion_matrix_button = ttk.Button(confusion_matrix_buttons_frame, text='State/County Model', command=self.show_fips_confusion_matrix)
+    lonlat_confusion_matrix_button = ttk.Button(confusion_matrix_buttons_frame, text='Longitude/Latitude Model', command=self.show_lonlat_confusion_matrix)
 
     # Set items on the grid
     title.grid(row=0, column=0, sticky=(N, E, W))
-    subframe.grid(row=1, column=0, sticky=(N, S, E, W))
-    input_frame.grid(row=0, column=0, sticky=NSEW)
+    subframe.grid(row=1, column=0, sticky=NSEW)
+    input_frame.grid(row=0, column=0, columnspan=2, sticky=NSEW)
     discovery_datetime_label.grid(row=0, column=0, sticky=E)
     self.discovery_datetime_entry.grid(row=0, column=1, sticky=W)
     contained_datetime_label.grid(row=1, column=0, sticky=E)
@@ -122,10 +125,9 @@ def __init__(self, *args, **kwargs):
     self.latitude_entry.grid(row=1, column=4, sticky=W)
     predict_button.grid(row=6, column=0, columnspan=4, pady=5, sticky=NSEW)
     display_frame.grid(row=1, column=0, sticky=NSEW)
-    display_frame_subtitle.grid(row=0, column=0, columnspan=3, sticky=(N, E, W))
-    predicted_category_label.grid(row=1, column=1, sticky=E)
-    predicted_category.grid(row=2, column=1, sticky=E)
-    probabilities_frame.grid(row=1, column=2, rowspan=4, sticky=(N, S, E))
+    predicted_category_label.grid(row=0, column=0, sticky=E)
+    predicted_category.grid(row=0, column=1, sticky=W)
+    probabilities_frame.grid(row=1, column=0, rowspan=2, columnspan=3, sticky=NSEW)
     probabilities_label.grid(row=0, column=0, columnspan=5, sticky=(N, E, W))
     probability_1_label.grid(row=1, column=0, sticky=E)
     probability_1_cause.grid(row=1, column=1, sticky=E)
@@ -142,27 +144,34 @@ def __init__(self, *args, **kwargs):
     probability_3_colon_label.grid(row=3, column=2)
     probability_3_percent.grid(row=3, column=3, sticky=W)
     probability_3_percent_label.grid(row=3, column=4, sticky=W)
+    confusion_matrix_buttons_frame.grid(row=1, column=1, sticky=NSEW)
+    fips_confusion_matrix_button.grid(row=0, column=0, sticky=NSEW, padx=4, pady=4)
+    lonlat_confusion_matrix_button.grid(row=1, column=0, sticky=NSEW, padx=4, pady=4)
 
     # Configure the grid
     self.rowconfigure((0,), weight=1)
     self.rowconfigure((1,), weight=5)
     self.columnconfigure((0,), weight=1)
     subframe.rowconfigure((0,), weight=2)
     subframe.rowconfigure((1,), weight=1)
-    subframe.columnconfigure((0,), weight=1)
+    subframe.columnconfigure((0,), weight=2)
+    subframe.columnconfigure((1,), weight=1)
     input_frame.rowconfigure((0, 1, 2, 3, 4, 5), weight=1)
     input_frame.rowconfigure((6,), weight=2)
     input_frame.columnconfigure((0, 1, 2, 3), weight=1)
     location_input_frame.rowconfigure((0, 1), weight=1)
     location_input_frame.columnconfigure((0, 1, 3, 4), weight=5)
     location_input_frame.columnconfigure((2,), weight=1)
     display_frame.rowconfigure((0,), weight=2)
-    display_frame.rowconfigure((1, 2, 3, 4), weight=1)
-    display_frame.columnconfigure((0, 1, 2), weight=1)
+    display_frame.rowconfigure((1, 2), weight=1)
+    display_frame.columnconfigure((0, 1), weight=1)
+    display_frame.columnconfigure((2,), weight=2)
     probabilities_frame.rowconfigure((0,), weight=3)
     probabilities_frame.rowconfigure((1, 2, 3), weight=2)
     probabilities_frame.columnconfigure((0, 2, 3, 4), weight=1)
     probabilities_frame.columnconfigure((1), weight=2)
+    confusion_matrix_buttons_frame.rowconfigure((0, 1), weight=1)
+    confusion_matrix_buttons_frame.columnconfigure((0,), weight=1)
 
   def set_up_state_county_variables(self):
     self.fips_codes_df = get_fips_codes_dataframe()
@@ -261,3 +270,9 @@ def display_predicted_info(self, probabilities):
     self.probability_1_percent_var.set(f'{probabilities[0][1] * 100:.2f}')
     self.probability_2_percent_var.set(f'{probabilities[1][1] * 100:.2f}')
     self.probability_3_percent_var.set(f'{probabilities[2][1] * 100:.2f}')
+
+  def show_fips_confusion_matrix(self):
+    plot_lonlat_confusion_matrix()
+
+  def show_lonlat_confusion_matrix(self):
+    plot_fipscode_confusion_matrix()

src/plotting.py

@@ -67,7 +67,7 @@ def plot_lonlat_model_confusion_matrix():
   df = fires_df[['fire_size', 'longitude', 'latitude', 'discovery_datetime', 'contained_datetime', 'stat_cause_code']].dropna()
   X = df.drop('stat_cause_code', axis=1)
   y = df['stat_cause_code'] - 1
-  fig, ax = plt.subplots(figsize=[12, 8])
+  fig, ax = plt.subplots(figsize=[10, 9])
   ConfusionMatrixDisplay.from_predictions(y_true=y, y_pred=model.predict(X), normalize='true', xticks_rotation='vertical', display_labels=STAT_CAUSE_CODE_TO_DESCR.values(), values_format='.2f', ax=ax)
   ax.set_title('Confusion Matrix for the Longitude/Latitude prediction model.')
   fig.tight_layout()
@@ -81,7 +81,7 @@ def plot_fipscode_model_confusion_matrix():
   df.loc[:, ['combined_fips_code']] = encoder.transform(df[['combined_fips_code']])[0].astype(int)
   X = df.drop('stat_cause_code', axis=1)
   y = df['stat_cause_code'] - 1
-  fig, ax = plt.subplots(figsize=[10, 10])
+  fig, ax = plt.subplots(figsize=[10, 9])
   ConfusionMatrixDisplay.from_predictions(y_true=y, y_pred=model.predict(X), normalize='true', xticks_rotation='vertical', display_labels=STAT_CAUSE_CODE_TO_DESCR.values(), values_format='.2f', ax=ax)
   ax.set_title('Confusion Matrix for the State/County prediction model.')
   fig.tight_layout()

src/prediction.py

@@ -12,7 +12,7 @@
   11: 'Powerline', 12: 'Structure', 13: 'Missing/Undefined'
 }
 
-def joblib_objects_unpacked():
+def joblib_objects_present():
   return FIPS_MODEL_PATH.is_file() and FIPS_ENCODER_PATH.is_file() and LONLAT_MODEL_PATH.is_file()
 
 _fips_encoder = None
@@ -48,7 +48,7 @@ def get_lonlat_model():
 def run_fips_model_prediction(fire_size, combined_fips_code, discovery_datetime, contained_datetime):
   encoder = get_fips_encoder()
   classifier = get_fips_model()
-  encoded_fips_code = encoder.transform([[combined_fips_code]])[0].astype(int)[0]
+  encoded_fips_code = encoder.transform(pd.DataFrame(data=[[combined_fips_code]], columns=['combined_fips_code']))[0].astype(int)[0]
   df = pd.DataFrame(
     data=[[fire_size, encoded_fips_code, discovery_datetime.timestamp(), contained_datetime.timestamp()]],
     columns=['fire_size', 'combined_fips_code', 'discovery_datetime', 'contained_datetime']