Simple, Multi-Layer, Recurrent, LSTM and Convolutional Neural Network with Tensorflow

3 min readDec 16, 2020

We are demonstaring simple, multi-layer, recurrent and LSTM neural network implementation with Tensorflow. The data file can be loaded from here. Check also github repository. It contains data from IT operations and we are considering it as binary classification problem (Normal vs some error). The first step is to load the data and do preprocessing in order to prepare the input to neural network. In this work we have applied neural nework frameworks for binary classification problem.

import pandas as pd 
import numpy as np# read a data file into the python console
df = pd.read_csv("D:\cloudaeye\python_part\data.csv")#To print all the features (columns)
count = 0;
for columns in df:
    count = count+1
    #print(columns)
print("number of features:", count)
print("number of feature instances", df.size)# We are labeling the status with NORMAL error as 1 and all other errors as 0
for i in range(0,df.status.size):
  if df.status[i] == 'NORMAL':
    df.status[i] = 1
  else: df.status[i] = 0# We are not considering timestamp
df = df.loc[:, df.columns != 'timestamp']# status column is fed as lebels in a seperate column
df_train = df.loc[:, df.columns != 'status']# create labels using status column
label = df.status
label=label.astype('int')#Split training and test data
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(df_train, label, test_size = 0.20, random_state = 42)

then you need to import tensorflow and keras libraries

from tensorflow import keras
from tensorflow.keras import layers, models
from keras.models import Sequential
from tensorflow.python import keras
from keras.layers.core import Dense
from keras.layers.recurrent import LSTM, SimpleRNN
from numpy import newaxis
import tensorflow as tf
import numpy as np

Simple Neural Network

following is the code for Simple Neural Network

#Simple Neural Network
model = keras.Sequential()
model.add(Dense(1, kernel_initializer = 'normal', input_shape = X_train.shape[1:]))
model.compile(loss= 'BinaryCrossentropy',optimizer='Adam', metrics=['accuracy'])
model.fit(X_train, y_train, epochs=10)

Multi-layer Neural Network

following is the code for Multi-Layer Neural Network

# Multi-layer Neural Network
model = keras.Sequential()
model.add(Dense(1, kernel_initializer = 'normal', input_shape = X_train.shape[1:] ))
model.add(Dense(1, kernel_initializer = 'normal', activation='softmax'))
model.add(Dense(1, activation='sigmoid'))
model.compile(loss= 'BinaryCrossentropy',optimizer='Adam', metrics=['accuracy'])
model.fit(X_train, y_train, epochs=10)

Input to recurrent neural network is a 3d array and what we have from the above preprocessing step is a 2d dataframe. Follow the steps to convert 2d dataframe to 3d array.

# conver 2d dataframe to 3d arrayX_train_ar = X_train.to_numpy() 
X_test_ar = X_test.to_numpy() X_train_3d = X_train_ar[:,newaxis,:]
X_test_3d = X_test_ar[:,newaxis,:]print(X_train_3d.shape)
print(X_test_3d.shape)print(y_train.shape)
print(y_test.shape)

Recurrent Neural Network

following is the code for Recurrent Neural Network

# RNN
model = Sequential()
model.add(SimpleRNN(5, input_shape=(X_train_3d.shape[1:]), return_sequences=True))
model.add(Dense(1, activation='softmax'))
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
model.fit(X_train_3d, y_train, epochs=30)

LSTM Neural Network

following is the code for LSTM Neural Network

# LSTM
model = Sequential()
model.add(LSTM(5, input_shape=(X_train_3d.shape[1:]), return_sequences=True))
model.add(Dense(1, activation='softmax'))
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])
model.fit(X_train_3d, y_train, epochs=30)

Convolutional Neural Network

Input to Convolutional neural network is a 4d array and what we have from the above preprocessing step is a 2d dataframe. Follow the steps to convert 2d dataframe to 4d array.

# convert 2d data frame to 4d array 
X_train_ar = X_train.to_numpy() 
X_test_ar = X_test.to_numpy()X_train_ar = np.float32(X_train_ar)
X_test_ar = np.float32(X_test_ar)res1 = np.concatenate([X_train_ar[np.newaxis] for arr in X_train_ar])
res2 = np.concatenate([X_test_ar[np.newaxis] for arr in X_test_ar])
X_train_4d = res1[:,:,:,newaxis]
X_test_4d = res2[:,:,:,newaxis]print(X_train_4d.shape)
print(X_test_4d.shape)print(y_train.shape)
print(y_test.shape)

following is the code for Convolutional Neural Network

model = keras.Sequential()
model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=X_train_4d.shape[1:]))
model.add(layers.MaxPooling2D((2, 2)))
model.add(layers.Conv2D(64, (3, 3), activation='relu'))model.add(layers.Flatten())
model.add(layers.Dense(64, activation='relu'))
model.add(layers.Dense(10))model.compile(optimizer='adam',
              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
              metrics=['accuracy'])
history = model.fit(X_train_4d, y_train, epochs=2)