ANN models with MLP sklearn

Author: cconsta1

.DS_Store

@@ -20,10 +20,10 @@
 import numpy as np
 import plotly.graph_objs as go
 from dash import Input, Output, State, dcc, html
-import tensorflow as tf
-from tensorflow import keras
-from keras.models import load_model
-import xgboost
+# import tensorflow as tf
+# from tensorflow import keras
+# from keras.models import load_model
+#import xgboost
 import re
 
 
@@ -462,7 +462,7 @@ def process_input(n_clicks1, n_clicks2, n_clicks3, n_clicks4, n_clicks5, n_click
 
             rmse, rmse_tf = regression_model_info_extractor("Suchey_Brooks_1990")
 
-            regression = output_regression(y_regression_sklearn[0],rmse,y_regression_tf[0][0],rmse_tf)
+            regression = output_regression(y_regression_sklearn[0],rmse,y_regression_tf[0],rmse_tf)
 
             classification = output_classification(y_classification_sklearn[0], \
                                                    y_classification_tf[0])
@@ -479,7 +479,7 @@ def process_input(n_clicks1, n_clicks2, n_clicks3, n_clicks4, n_clicks5, n_click
 
             rmse, rmse_tf = regression_model_info_extractor("Meindl_and_Lovejoy")
 
-            regression = output_regression(y_regression_sklearn[0],rmse,y_regression_tf[0][0],rmse_tf)
+            regression = output_regression(y_regression_sklearn[0],rmse,y_regression_tf[0],rmse_tf)
 
             classification = output_classification(y_classification_sklearn[0], \
                                                    y_classification_tf[0])
@@ -494,7 +494,7 @@ def process_input(n_clicks1, n_clicks2, n_clicks3, n_clicks4, n_clicks5, n_click
 
             rmse, rmse_tf = regression_model_info_extractor("Lovejoy_et_al")
 
-            regression = output_regression(y_regression_sklearn[0],rmse,y_regression_tf[0][0],rmse_tf)
+            regression = output_regression(y_regression_sklearn[0],rmse,y_regression_tf[0],rmse_tf)
 
             classification = output_classification(y_classification_sklearn[0], \
                                                    y_classification_tf[0])
@@ -509,7 +509,7 @@ def process_input(n_clicks1, n_clicks2, n_clicks3, n_clicks4, n_clicks5, n_click
 
             rmse, rmse_tf = regression_model_info_extractor("Buckberry_and_Chamberlain")
 
-            regression = output_regression(y_regression_sklearn[0],rmse,y_regression_tf[0][0],rmse_tf)
+            regression = output_regression(y_regression_sklearn[0],rmse,y_regression_tf[0],rmse_tf)
 
             classification = output_classification(y_classification_sklearn[0], \
                                                    y_classification_tf[0])
@@ -525,7 +525,7 @@ def process_input(n_clicks1, n_clicks2, n_clicks3, n_clicks4, n_clicks5, n_click
 
             rmse, rmse_tf = regression_model_info_extractor("Suchey_Brooks_1990_and_Lovejoy_et_al")
 
-            regression = output_regression(y_regression_sklearn[0],rmse,y_regression_tf[0][0],rmse_tf)
+            regression = output_regression(y_regression_sklearn[0],rmse,y_regression_tf[0],rmse_tf)
 
             classification = output_classification(y_classification_sklearn[0], \
                                                    y_classification_tf[0])
@@ -541,7 +541,7 @@ def process_input(n_clicks1, n_clicks2, n_clicks3, n_clicks4, n_clicks5, n_click
 
             rmse, rmse_tf = regression_model_info_extractor("Suchey_Brooks_1990_and_Buckberry_Chamberlain")
 
-            regression = output_regression(y_regression_sklearn[0],rmse,y_regression_tf[0][0],rmse_tf)
+            regression = output_regression(y_regression_sklearn[0],rmse,y_regression_tf[0],rmse_tf)
 
             classification = output_classification(y_classification_sklearn[0], \
                                                    y_classification_tf[0])
@@ -560,7 +560,7 @@ def process_input(n_clicks1, n_clicks2, n_clicks3, n_clicks4, n_clicks5, n_click
 
             rmse, rmse_tf = regression_model_info_extractor("All")
 
-            regression = output_regression(y_regression_sklearn[0],rmse,y_regression_tf[0][0],rmse_tf)
+            regression = output_regression(y_regression_sklearn[0],rmse,y_regression_tf[0],rmse_tf)
 
             classification = output_classification(y_classification_sklearn[0], \
                                                    y_classification_tf[0])
@@ -573,11 +573,11 @@ def process_input(n_clicks1, n_clicks2, n_clicks3, n_clicks4, n_clicks5, n_click
 def output_classification(y_sklearn, y_tf):
 
     text = (
-        f"The sample is split into three age-groups, "
-        f"14-34 (class 0), 35-49 (class 1), and 50- (class 2). "
-        f"Using the sklearn library's classification algorithms we predict that "
-        f"for the given input the sample belongs to the {y_sklearn} class, "
-        f"and using a tensorflow neural network the prediction for the class is {y_tf}."
+        f"The sample was divided into three age groups: "
+        f"14-34 (class 0), 35-49 (class 1), and 50+ (class 2). "
+        f"Utilizing classification algorithms from the sklearn library we predict "
+        f"that the given input belongs to class {y_sklearn}. "
+        f"Additionally, our neural network predicts that the input belongs to class {y_tf}."
     )
 
     card = dbc.Card(
@@ -599,12 +599,12 @@ def output_classification(y_sklearn, y_tf):
 def output_regression(result_sklearn, rmse_sklearn, result_tf, rmse_tf):
 
     text = (
-        f"Using regression, we can make a prediction for the age directly. "
-        f"Using the sklearn library's regression algorithms we predict an age of {result_sklearn:.1f}"
+        f"We can make a direct prediction for age using regression. "
+        f"With the help of regression algorithms from the sklearn library, we predict an age of {result_sklearn:.1f}"
         f"\u00B1"
         f"{rmse_sklearn:.1f}"
-        f", and "
-        f"using a tensorflow neural network we predict {result_tf:.1f}"
+        f". "
+        f"Additionally, our neural network predicts an age of {result_tf:.1f}"
         f"\u00B1"
         f"{rmse_tf:.1f}"
     )
@@ -630,20 +630,26 @@ def calculate_y_vectors(model, X):
         pickle.load(
             open("".join(["./models/classification_right_",model,".dat"]), "rb"))
     classification_model_tf = \
-        load_model(
-            "".join(["./models/ann_classification_right_",model,".h5"]))
+        pickle.load(
+            open("".join(["./models/ann_classification_right_",model,".dat"]), "rb"))
+        
+        
+    # load_model(
+    #         "".join(["./models/ann_classification_right_",model,".h5"]))
 
     regression_model_sklearn = \
         pickle.load(
             open("".join(["./models/regression_right_",model,".dat"]), "rb"))
     regression_model_tf = \
-        load_model(
-            "".join(["./models/ann_regression_right_",model,".h5"]))
+        pickle.load(
+            open("".join(["./models/ann_regression_right_",model,".dat"]), "rb"))
+        # load_model(
+        #     "".join(["./models/ann_regression_right_",model,".h5"]))
 
     y_classification_sklearn = classification_model_sklearn.predict(X)
 
     y_classification_tf = classification_model_tf.predict(X)
-    y_classification_tf = np.argmax(y_classification_tf, axis=1)
+    #y_classification_tf = np.argmax(y_classification_tf, axis=1)
 
     y_regression_sklearn = regression_model_sklearn.predict(X)
     y_regression_tf = regression_model_tf.predict(X)

app.py

@@ -1,9 +1,17 @@
 ---------------------------------
 All
 Dataset size: 140 140
-Accuracy: 0.6857142857142857
+Best classifier: {'learner': MLPClassifier(activation='tanh', alpha=0.004441038058541382,
+              beta_1=0.9527358121951149, beta_2=0.9799242066519736,
+              epsilon=9.89535978544575e-06, learning_rate='adaptive',
+              learning_rate_init=0.030775305216924835, max_fun=25541,
+              max_iter=273, momentum=0.9565810529873298, n_iter_no_change=20,
+              power_t=0.39730012597415304, random_state=4, solver='sgd',
+              tol=0.006128519233101691,
+              validation_fraction=0.18358166633762218), 'preprocs': (StandardScaler(with_std=False),), 'ex_preprocs': ()}
+Accuracy: 0.7142857142857143
 Confusion matrix: 
-[[ 5  2  0]
+[[ 6  1  0]
  [ 2  0  5]
  [ 0  2 19]]
-LOOCV accuracy: 0.7642857142857143
+20-fold accuracy: 0.7142857142857142

models/ann_classification_right_All.dat

@@ -1,9 +1,16 @@
 ---------------------------------
 Buckberry and Chamberlain
 Dataset size: 140 140
-Accuracy: 0.7428571428571429
+Best classifier: {'learner': MLPClassifier(alpha=0.0013798362489681195, beta_1=0.8427429717332454,
+              beta_2=0.9793499837952889, epsilon=6.782183444500067e-06,
+              hidden_layer_sizes=(20,), learning_rate='invscaling',
+              learning_rate_init=0.05362664460304082, max_fun=13641,
+              max_iter=259, momentum=0.9253775432649511, n_iter_no_change=30,
+              power_t=0.6932640248738308, random_state=0,
+              tol=0.00146402570692232, validation_fraction=0.07467194239359302), 'preprocs': (PCA(n_components=5),), 'ex_preprocs': ()}
+Accuracy: 0.7142857142857143
 Confusion matrix: 
-[[ 4  0  3]
- [ 0  1  6]
- [ 0  0 21]]
-LOOCV accuracy: 0.6928571428571428
+[[ 5  1  1]
+ [ 2  0  5]
+ [ 0  1 20]]
+20-fold accuracy: 0.6785714285714286

models/ann_classification_right_All.txt

@@ -1,9 +1,17 @@
 ---------------------------------
 Lovejoy et al
 Dataset size: 140 140
-Accuracy: 0.8
+Best classifier: {'learner': MLPClassifier(activation='identity', alpha=0.007888821884740403,
+              beta_1=0.9037988007792519, beta_2=0.9739918146106623,
+              epsilon=1.233568968982738e-06,
+              learning_rate_init=0.09178142853860238, max_fun=29535,
+              max_iter=270, momentum=0.9859029509207262, n_iter_no_change=20,
+              power_t=0.7946478078411516, random_state=3,
+              tol=0.009186083269769324,
+              validation_fraction=0.02832480065184169), 'preprocs': (MinMaxScaler(clip=True, feature_range=(0.0, 1.0)),), 'ex_preprocs': ()}
+Accuracy: 0.8285714285714286
 Confusion matrix: 
 [[ 7  0  0]
- [ 1  0  6]
+ [ 1  1  5]
  [ 0  0 21]]
-LOOCV accuracy: 0.7785714285714286
+20-fold accuracy: 0.7999999999999999

models/ann_classification_right_Buckberry_and_Chamberlain.dat

@@ -1,9 +1,17 @@
 ---------------------------------
 Meindl and Lovejoy
 Dataset size: 140 140
-Accuracy: 0.5428571428571428
+Best classifier: {'learner': MLPClassifier(activation='identity', alpha=0.0017893961349955139,
+              beta_1=0.8120661393066152, beta_2=0.9848215178951898,
+              epsilon=7.466114747734495e-06, hidden_layer_sizes=(10, 10),
+              learning_rate_init=0.048556142949039094, max_fun=20396,
+              max_iter=254, momentum=0.8455372036522589, n_iter_no_change=30,
+              power_t=0.7796736921110702, random_state=1, solver='sgd',
+              tol=0.007026905109325465,
+              validation_fraction=0.15163797733143308), 'preprocs': (StandardScaler(with_mean=False, with_std=False),), 'ex_preprocs': ()}
+Accuracy: 0.6
 Confusion matrix: 
 [[ 5  2  0]
  [ 0  1  6]
- [ 3  5 13]]
-LOOCV accuracy: 0.6571428571428571
+ [ 3  3 15]]
+20-fold accuracy: 0.6428571428571428