FOR WORKING APPLICATION

Project Details

Official Documentation

 

Key Points of Service Provided

This is a medical disease prediction project with the help of machine learning and artificial intelligence.This project is having both web and android versions.The link to the website is given below.It mainly focuses on the following diseases:-

  • Breast Cancer Prediction.
  • Heart Attack Prediction
  • Lower Spine Back Pain Disease Prediction .
  • Chronic Kidney Disease Prediction.
  • Diabeties Disease Prediction.
  • Liver Disease Prediction.
  • Lung Disease Prediction .

Used Technologies and Libraries in the Project:-

Technologies Used:-

Machine Learning,Artificial Neural Networks,Convulutional Neural Networks,Android Development,Web Development.

Requirements:-

Required GPU Support (Used Google Colab)

Description of Breast Cancer Prediction in Detail:-

  • Dataset is taken from kaggle.The link to the dataset is given here.
  • Here there are 2 types:
    • Cancer with 5 attributes.
    • Cancer with 8 attributes
  • The attributes used are
    • Smoothness Mean:- mean of local variation in radius lengths
    • Radius Mean:-mean of distances from center to points on the perimeter
    • Symmetry Mean
    • Fractral Dimension Mean:- Here Fractral Dimension refers to “coastline approximation” – 1
    • Radius Standard Error
    • Texture Standard Error:- standard error in standard deviation of gray-scale values
    • Smoothness Standard Error:- standard error in local variation in radius lengths
  • Here a Artificial Neural Network (ANN) is used. The Heat Map of the data correlation between features is shown below
  • Here the labels are Malignant and Benign
  • The dataset is originally taken from kaggle which is referenced to UCI Machine learning Database.
  • It contains two parts in both web and Android.
    • Cancer I (Which does Prediction with first 5 above mentioned attributes)
    • Cancer II ( Which Does Prediction with all above mentioned attributes)
  • The Dataset is performed with Exploratory Data Analysis and Need to remove null values from the dataset.
  • Now we need to clean the dataset from 32 attributes in original to 5/8 attributes respectively.
  • The model architectures for both the cases are described below.

Model Architecture for Cancer I

The first layer is input layer with input shape as (5,1) . The next layer is the hidden layer with 100 nodes and activation function as RELU. The next layer is output layer with dense layer consisting of 2 nodes with activation function as sigmoid.Now we need to compile the model using loss as binary cross entropy and optimizer as adam.Now the model is trained with 700 epochs and training accuracy is 90.33%.

The testing accuracy is 96.4%

Model Architecture for Cancer II

The first layer is input layer with input shape as (5,1) . The next layer is the hidden layer with 100 nodes and activation function as RELU. The next layer is output layer with dense layer consisting of 2 nodes with activation function as sigmoid.Now we need to compile the model using loss as binary cross entropy and optimizer as adam.Now the model is trained with 700 epochs and training accuracy is 90.33%.

The testing accuracy is 96.4%