From d49d890c3e6152434e7497e70e4319dcaac7ff98 Mon Sep 17 00:00:00 2001
From: Tereso del Rio <delriot@coventry.ac.uk>
Date: Mon, 31 Oct 2022 01:46:36 +0000
Subject: [PATCH] Modify my_mlp to improve perforemance

---
 basic_ml.py    | 10 +++++-----
 main.py        |  4 +---
 ml_results.csv |  7 +------
 3 files changed, 7 insertions(+), 14 deletions(-)

diff --git a/basic_ml.py b/basic_ml.py
index 85be808..941300f 100644
--- a/basic_ml.py
+++ b/basic_ml.py
@@ -53,17 +53,17 @@ def basic_ml(x_train, x_test, y_train, y_test, model, random_state=0):
     return accuracy_score(clf.predict(x_test), y_test)
 
 
-def use_tf(x_train, x_test, y_train, y_test, batch_size=64, epochs=100):
+def use_tf(x_train, x_test, y_train, y_test, batch_size=50, epochs=100):
     """Train a fully connected neural network and return its accuracy."""
     # design model
     model = keras.models.Sequential([
-        keras.layers.Dense(18, activation='relu',
+        keras.layers.Dense(80, activation='softmax',
                            input_shape=(x_train.shape[1],)),
-        keras.layers.Dense(6),
+        keras.layers.Dense(40, activation='softmax'),
+        keras.layers.Dense(6, activation='softmax'),
     ])
     # choose hyperparameters
-    print(y_train[:5])
-    loss = keras.losses.SparseCategoricalCrossentropy(from_logits=False)
+    loss = keras.losses.SparseCategoricalCrossentropy(from_logits=True)
     optim = keras.optimizers.Adam(learning_rate=0.001)
     metrics = ["accuracy"]
     model.compile(loss=loss, optimizer=optim, metrics=metrics)
diff --git a/main.py b/main.py
index 2d76a2d..a7407a7 100644
--- a/main.py
+++ b/main.py
@@ -18,10 +18,9 @@
 import pickle
 import random
 import csv
-from types import NoneType
 import importlib.util
 # Check if 'dataset_manipulation' is installed
-if importlib.util.find_spec('dataset_manipulation') is NoneType:
+if isinstance(importlib.util.find_spec('dataset_manipulation'), type(None)):
     from dataset_manipulation import name_unique_features
     from dataset_manipulation import remove_notunique_features
     from dataset_manipulation import balance_dataset
@@ -42,7 +41,6 @@
                                            'names_features_targets.txt')
 with open(names_features_targets_file, 'rb') as f:
     names, features, targets = pickle.load(f)
-print(len(features), len(features[0]), len(targets))
 augmented_features, augmented_targets = augmentate_dataset(features, targets)
 
 normalized_augmented_features = normalize(augmented_features)
diff --git a/ml_results.csv b/ml_results.csv
index 209bbd3..aede6aa 100644
--- a/ml_results.csv
+++ b/ml_results.csv
@@ -1,7 +1,2 @@
 Name,Normal,Balance data,Augment data

-SVC,0.21,0.21,0.16

-DT,0.32,0.27,0.34

-KNN,0.29,0.39,0.51

-RF,0.41,0.53,0.58

-MPL,0.21,0.2,0.17

-my_mlp,0.2,0.24,0.25

+my_mlp,0.28,0.31,0.41