GitHub - edno2819/Ransomware-Predicti: Desenvolvimento de Rede Neural de classificação de Ransoware para trabalho de Mestrado.

Previsão de classes de Ransoware

Nome da Familia ID

-----------------------------------------

Goodware            0

Critroni            1

CryptLocker         2

CryptoWall          3

KOLLAH              4

Kovter              5

Locker              6

MATSNU              7

PGPCODER            8

Reveton             9

TeslaCrypt         10

Trojan-Ransom      11

from keras.callbacks import EarlyStopping, ModelCheckpoint
from sklearn.model_selection import train_test_split
from sklearn.metrics import confusion_matrix
from sklearn.decomposition import PCA
from keras.models import Sequential
from sklearn import preprocessing
from keras.layers import Dense
from sklearn import metrics
import pandasql as ps
from utilsIA import *
import pandas as pd
import numpy as np
import itertools

Leitura do dataset

path = 'Ransoware.xlsx'
dataset = pd.read_excel(path)
dataset

	Ransomware	Classe	4	6	8	9	10	11	12	14	...	29079	29218	29287	29758	29769	29770	29796	29903	30200	30285
0	0	0	0	0	0	0	0	0	0	0	...	0	0	0	0	1	0	0	0	1	1
1	0	0	1	0	0	0	0	0	0	0	...	0	0	0	0	1	0	0	0	0	0
2	1	4	0	0	0	0	0	0	0	0	...	0	0	0	0	1	0	0	0	0	0
3	0	0	1	1	1	1	1	0	1	0	...	0	0	0	0	0	0	0	0	0	0
4	0	0	0	1	0	1	1	0	0	0	...	0	0	0	0	0	0	0	0	0	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
1519	1	4	0	1	0	0	0	0	0	0	...	0	0	0	0	1	0	0	0	1	1
1520	0	0	1	1	1	0	1	0	1	1	...	0	0	0	0	0	0	0	0	0	0
1521	0	0	1	1	1	0	0	0	0	0	...	0	0	0	0	1	0	1	0	0	0
1522	1	4	0	0	0	0	0	0	0	0	...	1	0	0	0	1	0	0	0	0	0
1523	0	0	1	1	1	0	1	0	1	0	...	0	0	0	0	0	0	0	0	0	0

1524 rows × 931 columns

Redução da dimensão do dataset usando PCA

Reduzindo as 929 colunas de modo que mantenha uma alta representatividade com os dados originais

train_row = 900
columns = 929
column_init = 2
column_label = 1

train_label = dataset.iloc[:train_row, column_label]
train_data = dataset.iloc[:train_row, column_init:columns+column_init]

test_label = dataset.iloc[train_row:, column_label].reset_index(drop=True)
test_data = dataset.iloc[train_row:, column_init:columns + column_init].reset_index(drop=True)

n_comp = 165
pca = PCA(n_components=n_comp)
X_train_pca = pca.fit_transform(train_data.values)
X_test_pca = pca.transform(test_data.values)
print(f'Representação: {pca.explained_variance_ratio_.sum()}')

Representação: 0.9715453985161124

X_train_pca = pd.DataFrame(X_train_pca)
X_test_pca = pd.DataFrame(X_test_pca)
df = pd.concat([X_train_pca, X_test_pca]).reset_index()
df['Classe']  = dataset['Classe']
df

	index	0	1	2	3	4	5	6	7	8	...	156	157	158	159	160	161	162	163	164	Classe
0	0	-4.821125	-0.487440	-2.201017	1.152447	-0.033623	1.352974	-0.578600	0.059564	0.038959	...	-0.053499	0.311921	-0.050478	-0.117299	0.066474	-0.203876	0.246825	-0.010021	-0.282411	0
1	1	-3.116892	-0.320086	-0.468679	0.077390	-0.733340	-0.059672	0.296803	-0.416835	-0.452211	...	-0.268512	0.186507	0.044608	0.237524	0.141487	-0.163366	-0.133308	-0.139559	0.031824	0
2	2	-4.817464	-0.726339	-2.223047	0.763878	0.408207	0.848962	-0.412907	0.157133	0.103711	...	0.115069	-0.030130	0.102367	0.148219	-0.342858	0.129300	-0.067159	-0.014187	0.124248	4
3	3	5.331648	-2.676628	1.265138	0.893076	1.153713	-0.109961	0.198922	-0.561980	-0.057770	...	0.036511	-0.175244	-0.323986	-0.024850	0.179646	-0.095332	-0.081275	-0.192800	0.104840	0
4	4	4.054361	1.525786	-0.874177	-0.684646	-0.014614	-0.708375	0.977450	0.126336	-0.524389	...	-0.058353	-0.232046	-0.426571	0.095166	-0.060333	-0.154573	-0.096402	-0.048692	0.468370	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
1519	619	-3.667744	-0.287221	-0.377622	-0.285508	-0.173065	-0.097861	-0.031989	-0.025919	0.053390	...	-0.033192	0.723736	-0.231849	-0.136856	-0.090843	0.277584	-0.120122	-0.174936	-0.394245	4
1520	620	1.207262	4.874456	3.782525	1.782034	0.379592	-0.096894	-0.562963	1.003049	-3.265090	...	-0.010782	-0.362701	-0.555329	0.299275	0.414574	0.270008	-0.037364	-0.070711	0.199022	0
1521	621	3.961603	1.436517	-1.136199	-0.921292	-0.807843	0.733954	1.622033	-2.877845	0.001690	...	-0.072145	-0.226455	0.478052	0.026402	0.172154	0.158334	-0.144294	-0.308613	-0.207756	0
1522	622	-4.470373	-0.806404	-1.547585	0.073598	0.697804	0.348658	-0.494289	0.044260	0.074465	...	-0.149596	-0.241636	0.059691	0.157608	0.282329	0.006757	-0.122828	-0.280303	-0.022284	4
1523	623	5.259862	-3.403548	0.995541	1.227109	0.961682	0.304587	-0.072982	-0.855472	0.098563	...	0.031373	0.019168	-0.094154	0.007101	-0.000693	0.014536	-0.059940	-0.031743	0.070975	0

1524 rows × 167 columns

Verificando a quantidade de cada Classe dentro do dataset

q1 = """
SELECT distinct count(*) as Quantidade,  Classe
FROM dataset 
group by Classe
order by Quantidade desc"""

ps.sqldf(q1)

	Quantidade	Classe
0	942	0
1	107	2
2	97	6
3	90	9
4	64	5
5	59	7
6	50	1
7	46	3
8	34	11
9	25	4
10	6	10
11	4	8

Exclusão das classes com poucas ocorrências

Treinamento insuficiente do modelo

Teste insuficiente do modelo

Evitar alta acurácia por aleatoriedade

df.drop(df.loc[df['Classe']==8].index, inplace=True)
df.drop(df.loc[df['Classe']==10].index, inplace=True)
df.drop(df.loc[df['Classe']==4].index, inplace=True)
df.drop('index', 1, inplace=True)
df

C:\Users\Edno\AppData\Local\Temp/ipykernel_5244/4213288135.py:4: FutureWarning: In a future version of pandas all arguments of DataFrame.drop except for the argument 'labels' will be keyword-only
  df.drop('index', 1, inplace=True)

	0	1	2	3	4	5	6	7	8	9	...	156	157	158	159	160	161	162	163	164	Classe
0	-4.821125	-0.487440	-2.201017	1.152447	-0.033623	1.352974	-0.578600	0.059564	0.038959	0.154728	...	-0.053499	0.311921	-0.050478	-0.117299	0.066474	-0.203876	0.246825	-0.010021	-0.282411	0
1	-3.116892	-0.320086	-0.468679	0.077390	-0.733340	-0.059672	0.296803	-0.416835	-0.452211	0.099741	...	-0.268512	0.186507	0.044608	0.237524	0.141487	-0.163366	-0.133308	-0.139559	0.031824	0
3	5.331648	-2.676628	1.265138	0.893076	1.153713	-0.109961	0.198922	-0.561980	-0.057770	1.013530	...	0.036511	-0.175244	-0.323986	-0.024850	0.179646	-0.095332	-0.081275	-0.192800	0.104840	0
4	4.054361	1.525786	-0.874177	-0.684646	-0.014614	-0.708375	0.977450	0.126336	-0.524389	-0.025725	...	-0.058353	-0.232046	-0.426571	0.095166	-0.060333	-0.154573	-0.096402	-0.048692	0.468370	0
5	1.761522	1.777536	1.389631	-0.734398	-1.769632	-1.707925	0.010637	1.006003	-1.032105	0.677926	...	-0.066147	-0.053367	-0.184165	-0.156341	0.059520	-0.043059	-0.013471	-0.092569	0.110692	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
1516	5.471322	1.039605	-4.132083	0.309695	-2.036437	0.087284	0.457468	-0.841350	-0.999483	-1.548910	...	0.175220	-0.015662	-0.196201	-0.293100	-0.004059	-0.188156	0.085008	0.043024	0.112271	0
1517	5.177134	-3.400939	1.034620	1.240165	0.937027	0.308304	-0.070532	-0.924902	0.166870	1.175099	...	0.038146	-0.034900	-0.006361	-0.103692	-0.046950	0.016334	-0.074340	-0.039927	-0.009890	0
1520	1.207262	4.874456	3.782525	1.782034	0.379592	-0.096894	-0.562963	1.003049	-3.265090	4.992138	...	-0.010782	-0.362701	-0.555329	0.299275	0.414574	0.270008	-0.037364	-0.070711	0.199022	0
1521	3.961603	1.436517	-1.136199	-0.921292	-0.807843	0.733954	1.622033	-2.877845	0.001690	2.131293	...	-0.072145	-0.226455	0.478052	0.026402	0.172154	0.158334	-0.144294	-0.308613	-0.207756	0
1523	5.259862	-3.403548	0.995541	1.227109	0.961682	0.304587	-0.072982	-0.855472	0.098563	1.115860	...	0.031373	0.019168	-0.094154	0.007101	-0.000693	0.014536	-0.059940	-0.031743	0.070975	0

1489 rows × 166 columns

q1 = """
SELECT distinct count(*) as Quantidade,  Classe
FROM df 
group by Classe
order by Quantidade desc"""

ps.sqldf(q1)

	Quantidade	Classe
0	942	0
1	107	2
2	97	6
3	90	9
4	64	5
5	59	7
6	50	1
7	46	3
8	34	11

Redução da quantidade de registros da categoria '0' para redução da chance de
enviesamento do modelo para esta classe

df_balanced = ProcessingData.balancedData(df, 'Classe', 107)
#df_balanced = df

ProcessingData.showLabelsQtd(df_balanced, 'Classe')

Quantidade de itens nas 9 categorias:
0     107
2     107
6      97
9      90
5      64
7      59
1      50
3      46
11     34
Name: Classe, dtype: int64

#Colocando a coluna 'Classe' na primeira posição
col = ['Classe'] + [c for c in range(n_comp)]
df_balanced = df_balanced[col]
df_balanced.head()

	0	1	2	3	4	5	6	7	8	...	155	156	157	158	159	160	161	162	163	164
1112	5.050851	-3.437326	0.880903	1.284616	1.039539	0.332499	-0.112654	-0.929544	0.185957	...	-0.003749	0.007185	-0.091705	0.030639	-0.184883	-0.014107	0.023129	-0.065869	-0.118227	0.052917
904	-4.744892	-1.090630	-1.894263	0.623459	0.556025	0.847031	-0.458654	0.282898	0.140678	...	-0.006809	0.061832	0.089955	-0.009707	0.050636	-0.008242	0.017653	0.002273	0.099682	-0.081503
1018	-0.894416	1.380315	0.817547	0.487481	-3.652643	-0.308384	0.959425	0.111086	-0.468956	...	0.020760	-0.028250	0.168435	-0.041256	-0.425442	0.265058	0.060733	0.246020	0.026194	-0.033660
294	5.585781	-3.418598	1.011767	1.184437	1.192173	0.112080	-0.042108	-0.783191	-0.261409	...	0.015675	-0.016180	-0.023752	0.092484	-0.102187	-0.064161	0.023080	0.000156	-0.029936	-0.090316
1355	4.773488	0.741391	-3.426230	0.476568	-1.845094	-0.454420	1.943682	-1.763280	-0.361983	...	-0.238522	-0.164905	0.178833	0.062664	0.089606	-0.098259	0.022611	0.108016	-0.008297	0.028628

5 rows × 166 columns

Aplicação do One Hot Encode

n_classes = len(df_balanced['Classe'].astype(int).value_counts())

datas = df_balanced.iloc[:, 1:].values

enc = preprocessing.OneHotEncoder()
labels = df_balanced.Classe.to_list()
labels = [[la] for la in labels]
enc.fit(labels)
labels_hot = enc.transform(labels).toarray()

Separando os dados para teste e treino

X_train, X_test, y_train, y_test = train_test_split(datas, labels_hot, test_size=.3, random_state=0)#, stratify=labels)
y_train = np.array(y_train) 
y_test = np.array(y_test)

Modelo da rede neural utilizada

model = Sequential()
model.add(Dense(380, activation='relu', input_dim=n_comp))
model.add(Dense(160, activation='relu'))
model.add(Dense(90, activation='relu'))
model.add(Dense(45, activation='relu'))
model.add(Dense(n_classes, activation='softmax'))


model.compile(optimizer='adam', loss='categorical_crossentropy', metrics = ['accuracy'])

callbacks = [EarlyStopping(monitor='val_accuracy', patience=25),
        ModelCheckpoint(filepath='best_model.h5', monitor='val_accuracy', save_best_only=True)]



history = model.fit(X_train, y_train, epochs=200, callbacks=callbacks, validation_data=(X_test, y_test), verbose=1)
model.load_weights('best_model.h5')

Epoch 1/200
15/15 [==============================] - 1s 23ms/step - loss: 1.9844 - accuracy: 0.2910 - val_loss: 1.7298 - val_accuracy: 0.4162
Epoch 2/200
15/15 [==============================] - 0s 9ms/step - loss: 1.4077 - accuracy: 0.6105 - val_loss: 1.2256 - val_accuracy: 0.5888
Epoch 3/200
15/15 [==============================] - 0s 10ms/step - loss: 0.9652 - accuracy: 0.7068 - val_loss: 1.0074 - val_accuracy: 0.6853
Epoch 4/200
15/15 [==============================] - 0s 10ms/step - loss: 0.7094 - accuracy: 0.7746 - val_loss: 0.8882 - val_accuracy: 0.7157
Epoch 5/200
15/15 [==============================] - 0s 9ms/step - loss: 0.5448 - accuracy: 0.8184 - val_loss: 0.7992 - val_accuracy: 0.7665
Epoch 6/200
15/15 [==============================] - 0s 9ms/step - loss: 0.3985 - accuracy: 0.8884 - val_loss: 0.7419 - val_accuracy: 0.7716
Epoch 7/200
15/15 [==============================] - 0s 5ms/step - loss: 0.3295 - accuracy: 0.9015 - val_loss: 0.7838 - val_accuracy: 0.7513
Epoch 8/200
15/15 [==============================] - 0s 9ms/step - loss: 0.2932 - accuracy: 0.9103 - val_loss: 0.7610 - val_accuracy: 0.8122
Epoch 9/200
15/15 [==============================] - 0s 5ms/step - loss: 0.2505 - accuracy: 0.9212 - val_loss: 0.7309 - val_accuracy: 0.7766
Epoch 10/200
15/15 [==============================] - 0s 9ms/step - loss: 0.2223 - accuracy: 0.9344 - val_loss: 0.7284 - val_accuracy: 0.8173
Epoch 11/200
15/15 [==============================] - 0s 9ms/step - loss: 0.1786 - accuracy: 0.9475 - val_loss: 0.7725 - val_accuracy: 0.8223
Epoch 12/200
15/15 [==============================] - 0s 6ms/step - loss: 0.1751 - accuracy: 0.9453 - val_loss: 0.7342 - val_accuracy: 0.8020
Epoch 13/200
15/15 [==============================] - 0s 9ms/step - loss: 0.1385 - accuracy: 0.9562 - val_loss: 0.7827 - val_accuracy: 0.8325
Epoch 14/200
15/15 [==============================] - 0s 5ms/step - loss: 0.1194 - accuracy: 0.9606 - val_loss: 0.7634 - val_accuracy: 0.8274
Epoch 15/200
15/15 [==============================] - 0s 5ms/step - loss: 0.1201 - accuracy: 0.9628 - val_loss: 0.7973 - val_accuracy: 0.8173
Epoch 16/200
15/15 [==============================] - 0s 5ms/step - loss: 0.1045 - accuracy: 0.9606 - val_loss: 0.8194 - val_accuracy: 0.8223
Epoch 17/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0962 - accuracy: 0.9781 - val_loss: 0.8739 - val_accuracy: 0.8223
Epoch 18/200
15/15 [==============================] - 0s 5ms/step - loss: 0.1099 - accuracy: 0.9628 - val_loss: 0.8371 - val_accuracy: 0.8325
Epoch 19/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0886 - accuracy: 0.9759 - val_loss: 0.8321 - val_accuracy: 0.8223
Epoch 20/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0760 - accuracy: 0.9759 - val_loss: 0.8440 - val_accuracy: 0.8325
Epoch 21/200
15/15 [==============================] - 0s 9ms/step - loss: 0.0788 - accuracy: 0.9803 - val_loss: 0.8539 - val_accuracy: 0.8376
Epoch 22/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0632 - accuracy: 0.9825 - val_loss: 0.8903 - val_accuracy: 0.8173
Epoch 23/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0817 - accuracy: 0.9716 - val_loss: 0.8795 - val_accuracy: 0.8274
Epoch 24/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0654 - accuracy: 0.9825 - val_loss: 0.9905 - val_accuracy: 0.8274
Epoch 25/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0876 - accuracy: 0.9716 - val_loss: 0.8827 - val_accuracy: 0.8325
Epoch 26/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0677 - accuracy: 0.9825 - val_loss: 0.9454 - val_accuracy: 0.8173
Epoch 27/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0775 - accuracy: 0.9759 - val_loss: 0.9085 - val_accuracy: 0.8274
Epoch 28/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0712 - accuracy: 0.9781 - val_loss: 0.9352 - val_accuracy: 0.8325
Epoch 29/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0671 - accuracy: 0.9825 - val_loss: 0.9524 - val_accuracy: 0.8122
Epoch 30/200
15/15 [==============================] - 0s 6ms/step - loss: 0.0605 - accuracy: 0.9825 - val_loss: 0.9559 - val_accuracy: 0.8223
Epoch 31/200
15/15 [==============================] - 0s 7ms/step - loss: 0.0775 - accuracy: 0.9781 - val_loss: 0.9911 - val_accuracy: 0.8274
Epoch 32/200
15/15 [==============================] - 0s 6ms/step - loss: 0.0886 - accuracy: 0.9716 - val_loss: 0.9290 - val_accuracy: 0.8325
Epoch 33/200
15/15 [==============================] - 0s 7ms/step - loss: 0.0729 - accuracy: 0.9803 - val_loss: 0.9835 - val_accuracy: 0.8274
Epoch 34/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0556 - accuracy: 0.9847 - val_loss: 0.9801 - val_accuracy: 0.8325
Epoch 35/200
15/15 [==============================] - 0s 6ms/step - loss: 0.0586 - accuracy: 0.9825 - val_loss: 0.9715 - val_accuracy: 0.8173
Epoch 36/200
15/15 [==============================] - 0s 6ms/step - loss: 0.0566 - accuracy: 0.9803 - val_loss: 0.9987 - val_accuracy: 0.8325
Epoch 37/200
15/15 [==============================] - 0s 7ms/step - loss: 0.0683 - accuracy: 0.9781 - val_loss: 0.9747 - val_accuracy: 0.8173
Epoch 38/200
15/15 [==============================] - 0s 8ms/step - loss: 0.0949 - accuracy: 0.9716 - val_loss: 1.0825 - val_accuracy: 0.8274
Epoch 39/200
15/15 [==============================] - 0s 7ms/step - loss: 0.0746 - accuracy: 0.9672 - val_loss: 0.9609 - val_accuracy: 0.8325
Epoch 40/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0505 - accuracy: 0.9847 - val_loss: 1.0257 - val_accuracy: 0.8223
Epoch 41/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0544 - accuracy: 0.9847 - val_loss: 0.9983 - val_accuracy: 0.8376
Epoch 42/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0465 - accuracy: 0.9847 - val_loss: 1.0335 - val_accuracy: 0.8223
Epoch 43/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0485 - accuracy: 0.9847 - val_loss: 1.0488 - val_accuracy: 0.8223
Epoch 44/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0454 - accuracy: 0.9847 - val_loss: 1.0534 - val_accuracy: 0.8376
Epoch 45/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0509 - accuracy: 0.9803 - val_loss: 1.0727 - val_accuracy: 0.8274
Epoch 46/200
15/15 [==============================] - 0s 5ms/step - loss: 0.0494 - accuracy: 0.9847 - val_loss: 1.0572 - val_accuracy: 0.8274

Avaliação do modelo

Histórico da evolução do modelo e sua Loss function

Acurácia do modelo nos dados de teste

Matriz de confução

def plot_confusion_matrix(y_test, y_pred, labels, normalize=False):
    cm = confusion_matrix(y_test, y_pred)

    if normalize:
        cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]
        print("Normalized confusion matrix")
    else:
        print('Confusion matrix, without normalization')

    plt.imshow(cm, interpolation='nearest', cmap=plt.cm.Blues)
    plt.gcf().set_size_inches(17, 11)
    plt.title('Confusion matrix')
    plt.colorbar()
    tick_marks = np.arange(len(labels))
    plt.xticks(tick_marks, labels, rotation=45)
    plt.yticks(tick_marks, labels)

    fmt = '.2f' if normalize else 'd'
    thresh = cm.max() / 2.
    for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):
        plt.text(j, i, format(cm[i, j], fmt),
                 horizontalalignment="center",
                 color="white" if cm[i, j] > thresh else "black")

    plt.tight_layout()
    plt.ylabel('True label')
    plt.xlabel('Predicted label')
    plt.show()

def evaluate_model(model, history, X_test, y_test):
    scores = model.evaluate((X_test), y_test, verbose=0)
    print('=========================================')
    print("|| Accuracy: %.2f%%" % (scores[1]*100))
    print('=========================================')

    fig, axs = plt.subplots(1, 2, figsize=(12,6))
    axs[0].plot(history.history['accuracy'])
    axs[0].plot(history.history['val_accuracy'])
    axs[0].set_title("Accuracy")
    axs[0].legend(['Training', 'Validation'])
    axs[1].plot(history.history['loss'])
    axs[1].plot(history.history['val_loss'])
    axs[1].set_title("Model- Loss")
    axs[1].legend(['Training', 'Validation'])
    fig.tight_layout()

evaluate_model(model, history, X_test, y_test)

=========================================
|| Accuracy: 83.76%
=========================================

OBS: O valores numéricos das classes não correspondem mais com os apresentados no inicio do arquivo

def find_class(probl):
    probl_max = probl.max()
    label = int(np.where(probl == probl_max)[0])

    return  label, round(probl_max*100,1)

prediction_proba = model.predict(X_test)

prediction_cat = [find_class(c)[0] for c in prediction_proba]
y_test_label =  [find_class(c)[0] for c in y_test]


plot_confusion_matrix(y_test_label, prediction_cat, enc.categories_[0], True)

Normalized confusion matrix

Área Sob a Curva ROC

Geralmente, a sensibilidade e a especificidade são características difíceis de conciliar, isto é, é complicado aumentar
a sensibilidade e a especificidade de um teste ao mesmo tempo. As curvas ROC (receiver operator characteristic curve)
são uma forma de representar a relação, normalmente antagónica, entre a sensibilidade e a especificidade de um teste
diagnóstico quantitativo, ao longo de um contínuo de valores de "cutoff point".

# First aggregate all false positive rates
all_fpr = np.unique(np.concatenate([fpr[i] for i in range(n_classes)]))

# Then interpolate all ROC curves at this points
mean_tpr = np.zeros_like(all_fpr)
for i in range(n_classes):
    mean_tpr += np.interp(all_fpr, fpr[i], tpr[i])

# Finally average it and compute AUC
mean_tpr /= n_classes

fpr["macro"] = all_fpr
tpr["macro"] = mean_tpr
roc_auc["macro"] = metrics.auc(fpr["macro"], tpr["macro"])

# Plot all ROC curves
plt.figure(figsize=(12,8))
plt.plot(
    fpr["micro"],
    tpr["micro"],
    label="Média-mínima da curva ROC (area = {0:0.2f})".format(roc_auc["micro"]),
    color="deeppink",
    linestyle=":",
    linewidth=4,
)

plt.plot(
    fpr["macro"],
    tpr["macro"],
    label="Média-máxima da curva ROC(area = {0:0.2f})".format(roc_auc["macro"]),
    color="navy",
    linestyle=":",
    linewidth=4,
)

for i in range(n_classes):
    plt.plot(
        fpr[i],
        tpr[i],
        lw=lw,
        label="Cruva ROC da classe {0} (area = {1:0.2f})".format(i, roc_auc[i]),
    )

plt.plot([0, 1], [0, 1], "k--", lw=lw)
plt.xlim([0.0, 1.0])
plt.ylim([0.0, 1.05])
plt.xlabel("Taxa de Falso Positivo")
plt.ylabel("Taxa de Verdadeiro Positivo")
plt.title("Curva ROC por classes")
plt.legend(loc="lower right")
plt.show()

Name		Name	Last commit message	Last commit date
Latest commit History 6 Commits
infos Dataset		infos Dataset
main_files		main_files
.gitignore		.gitignore
README.md		README.md
Ransoware.xlsx		Ransoware.xlsx
best_model.h5		best_model.h5
main.ipynb		main.ipynb
ransomware.csv		ransomware.csv
utilsIA.py		utilsIA.py

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

Redução da dimensão do dataset usando PCA

Exclusão das classes com poucas ocorrências

Separando os dados para teste e treino
X_train, X_test, y_train, y_test = train_test_split(datas, labels_hot, test_size=.3, random_state=0)#, stratify=labels) y_train = np.array(y_train) y_test = np.array(y_test)

Avaliação do modelo

Área Sob a Curva ROC

About

Releases

Packages

Languages

edno2819/Ransomware-Predicti

Folders and files

Latest commit

History

Repository files navigation

Redução da dimensão do dataset usando PCA

Verificando a quantidade de cada Classe dentro do dataset q1 = """ SELECT distinct count(*) as Quantidade, Classe FROM dataset group by Classe order by Quantidade desc""" ps.sqldf(q1) Quantidade Classe 0 942 0 1 107 2 2 97 6 3 90 9 4 64 5 5 59 7 6 50 1 7 46 3 8 34 11 9 25 4 10 6 10 11 4 8

Exclusão das classes com poucas ocorrências

Separando os dados para teste e treino X_train, X_test, y_train, y_test = train_test_split(datas, labels_hot, test_size=.3, random_state=0)#, stratify=labels) y_train = np.array(y_train) y_test = np.array(y_test)

Avaliação do modelo

Área Sob a Curva ROC

About

Topics

Resources

Stars

Watchers

Forks

Releases

Packages 0

Languages

Separando os dados para teste e treino
X_train, X_test, y_train, y_test = train_test_split(datas, labels_hot, test_size=.3, random_state=0)#, stratify=labels) y_train = np.array(y_train) y_test = np.array(y_test)

Packages