Dr.Diabetes
Using Advanced Machine Learning Algorithms and Neural Networks to predict diabetes in a patient with absolutely no use of medical testing
Revolutionizing The Medical Industry Using Machine Learning
?
More than 30 million people in the United States have diabetes, and 1 in 4 of them don’t know they have it.
Diabetes is the 7th leading cause of death in the United States
$
Medical costs and lost work and wages for people with Diagnosed Diabetes total $327 Billion Yearly
The prevalence of diagnosed diabetes in the U.S. increased by 382% from 1988 to 2014
Our Solution
Dr. Diabetes
Using Advanced Machine Learning and an Artificial Neural Network our team was able to build a powerful python script that would be able to determine whether or not an individual has diabetes based on a variety of preexisting health data
With this technology, the computer automatically learns all on its own with no direct instructions on how to identify the disease. Using previously compiled data sets, the computer is able to determine with great accuracy whether you have diabetes or not
Since we are using raw computation to determine a patient's condition, our solution will eliminate the need for extensive medical testing, allow people to diagnose their condition before it becomes critical, and ultimately save countless lives and expenses.
Scroll Down To Learn More
Core
Technology
01
02
Machine Learning
Artificial Neural Networking
Machine learning is a subset of artificial intelligence in the field of computer science that often uses statistical techniques to give computers the ability to "learn" with data, without being explicitly programmed.
A computing system made up of a number of simple, highly connected processing elements, which process inputs - Robert Nielson
Example:
The key to machine learning is letting the computer make its own decisions.
For example, if you took a picture of a bird and wanted to determine its breed using a computer, You could directly create a decision tree based on attributes:
if (color == red && wingspan <= 5)
bird = Cardinal
However, this can get extremely complicated with hundreds of features operating simultaneously.
Instead we feed the computer a set of training data and use pattern recognition algorithms to allow the computer to make its own criteria for what defines a Cardinal
Steps To Machine learning
Data Extraction
1
Model Selection
3
Parameter Tuning
5
Data Optimiziation
2
Training/Evaluation
4
Final Prediction
6
Training/Ev
Scroll down for how we used these in our code
Data Extraction
In our project, we used an open-source research database that gave us information on 1,000 individuals. Each individual was characterized by a 1 or 0 ( 1 = diabetes, 0 = healthy) along with 7 key health factors: Number of times pregnant, Glucose Concentration, Blood Pressure, Skin Thickness, Insulin Concentration, Body Mass Index, and Age
Data Optimization
Our goal is to use this data as a training set for our computer so it can use its pattern recognition algorithm to predict the diabetic condition of a brand new patient. However, before doing this we must normalize all the data points and remove missing data to prevent outliers and inaccuracy.
Model Selection
There are many models, mathematical systems of training a computer, in machine learning. In this scenario, our team used a sequential model as it is very efficient to use in binary classification(diabetes or not) and supports the addition of a neural network.
Training/Evaluation
The core element of machine learning is to train our computer to identify the correct solution by itself. In order to do this, we must carry tests and evaluate their accuracy. We trained the data using our layers built in our sequential model and used a 3-Fold Classification algorithm using a Grid Search to cross-validate our data(check if right or wrong)
Parametric Tuning
After we train/evaluate the data, we will want to further improve our accuracy by optimizing the parameters in our model. For example, in our case we optimized for the number of inputs the model runs through the data set(batch size/epochs), learning rate/dropout rate, and the number of active neurons. All of these will help improve the accuracy of our prediction.
Final Prediction
The whole point of a machine learning algorithm is to make a final prediction for an entirely new datapoint. Here we used a new datapoint with the 8 health inputs and were able to predict that yes, this person does in fact have diabetes.
This was done with no official medical testing, just patterns and computer.
How cool is that!
STAY IN TOUCH
Vinodh Sekhar
Jayachandra Ruthala
224-392-0254
?
More than 30 million people in the United States have diabetes, and 1 in 4 of them don’t know they have it.
Diabetes is the 7th leading cause of death in the United States
$
Medical costs and lost work and wages for people with Diagnosed Diabetes total $327 Billion Yearly
The prevalence of diagnosed diabetes in the U.S. increased by 382% from 1988 to 2014
Our Solution
Dr. Diabetes
Using Advanced Machine Learning and an Artificial Neural Network our team was able to build a powerful python script that would be able to determine whether or not an individual has diabetes based on a variety of preexisting health data
With this technology, the computer automatically learns all on its own with no direct instructions on how to identify the disease. Using previously compiled data sets, the computer is able to determine with great accuracy whether you have diabetes or not
Since we are using raw computation to determine a patient's condition, our solution will eliminate the need for extensive medical testing, allow people to diagnose their condition before it becomes critical, and ultimately save countless lives and expenses.
Scroll Down To Learn More
Core
Technology
01
02
Machine Learning
Artificial Neural Networking
Machine learning is a subset of artificial intelligence in the field of computer science that often uses statistical techniques to give computers the ability to "learn" with data, without being explicitly programmed.
A computing system made up of a number of simple, highly connected processing elements, which process inputs - Robert Nielson
Example:
The key to machine learning is letting the computer make its own decisions.
For example, if you took a picture of a bird and wanted to determine its breed using a computer, You could directly create a decision tree based on attributes:
if (color == red && wingspan <= 5)
bird = Cardinal
However, this can get extremely complicated with hundreds of features operating simultaneously.
Instead we feed the computer a set of training data and use pattern recognition algorithms to allow the computer to make its own criteria for what defines a Cardinal
Steps To Machine learning
Data Extraction
1
Model Selection
3
Parameter Tuning
5
Data Optimiziation
2
Training/Evaluation
4
Final Prediction
6
Scroll down for how we used these in our code
Data Extraction
In our project, we used an open-source research database that gave us information on 1,000 individuals. Each individual was characterized by a 1 or 0 ( 1 = diabetes, 0 = healthy) along with 7 key health factors: Number of times pregnant, Glucose Concentration, Blood Pressure, Skin Thickness, Insulin Concentration, Body Mass Index, and Age
Data Optimization
Our goal is to use this data as a training set for our computer so it can use its pattern recognition algorithm to predict the diabetic condition of a brand new patient. However, before doing this we must normalize all the data points and remove missing data to prevent outliers and inaccuracy.
Model Selection
There are many models, mathematical systems of training a computer, in machine learning. In this scenario, our team used a sequential model as it is very efficient to use in binary classification(diabetes or not) and supports the addition of a neural network.
Training/Evaluation
The core element of machine learning is to train our computer to identify the correct solution by itself. In order to do this, we must carry tests and evaluate their accuracy. We trained the data using our layers built in our sequential model and used a 3-Fold Classification algorithm using a Grid Search to cross-validate our data(check if right or wrong)
Parametric Tuning
After we train/evaluate the data, we will want to further improve our accuracy by optimizing the parameters in our model. For example, in our case we optimized for the number of inputs the model runs through the data set(batch size/epochs), learning rate/dropout rate, and the number of active neurons. All of these will help improve the accuracy of our prediction.
Final Prediction
The whole point of a machine learning algorithm is to make a final prediction for an entirely new datapoint. Here we used a new datapoint with the 8 health inputs and were able to predict that yes, this person does in fact have diabetes.
This was done with no official medical testing, just patterns and computer.
How cool is that!
Neural Network
The Neural Network is the main system we use to implement our models. They work by using an algorithm called backpropagation to generate predictive output. These predictions are compared to the actual results and the parameters are then re-modified. This cycle continues thousands of times to ensure optimal accuracy
Watch our Project in Action
© 2023 by Polystat. Proudly created with Wix.com