Build the Artificial Intelligence for detecting diabetes using Neural Networks and keras.

One approach focused on biological processes in the brain while the other focused on the application of neural networks to artificial intelligence.Simple Neural NetworkThe simple neural network is a mathematical model which simply takes Xn as inputs, multiply inputs with weights Wn and add bias b and pass it from activation function and returns Y as output as shown in below..Single layer simple neural network also called as perceptron.From the above diagram, the activation function is nothing but nonlinear function like sigmoid and relu to achieve nonlinearity..For example, all negative values are assigned to zero in sigmoid function..The purpose of using activation function in the neural network to get the smooth loss function means loss function has to be a convex function or concave function then we can apply gradient descent on it..We will talk about that ahead while implementation.python ImplementationIt is easy to implement the neural network in python using keras library..Keras is a python library built on top of tensorflow which is a very powerful tool for performing deep learning developed by Google..We will use Pima diabetes dataset and built a neural network which will predict if a patient is diabetic or not given inputs X like Glucose, blood pressure, skin thickness, etc..First, install keras and tensorflow using pip..For full code and data click here.Now we will load the data using pandas and visualize first five rows or records.Then we will choose dependent variable y and independent variables X..After choosing X and y we will scale the data and split them into training and testing sets respectively..Also, we will do one-hot encoding to our target variable y.After this, we will move towards designing and building the neural network model using keras..Here we are using one input layer with 12 neurons, one hidden layer with 8 neurons and one output layer with 2 neurons.. More details

Leave a Reply