Inspiration
Autism Spectrum Disorder is a disease that needs to be caught early on in order for the patient at risk to lead a normal, or near normal life. Research shows that self reported surveys and questionnaires are effective in determining the risk of autism in children, but such tests are expensive and time consuming.
What it does
The program(in its current state) generates a random data set that models the list of questionnaire responses for a particular child. It is then trained to determine which children are likely to develop autism using part of this data set. Randomly generated data was used because actual data requires the signing off of a private investigator associated with a university.
How I built it
The program was built using python, and makes extensive use of Pandas, Numplot and Sklearn in order to manipulate data and set up the machine learning algorithm.
Challenges I ran into
Organizing the data efficiently required the use of arrays and a data frame, both of which had different methods to splice and modify data. Figuring out the proper methods to analyze and process the data correctly required a lot of trial and error.
Accomplishments that I'm proud of
I was able to build something by the end of hackMIT, and managed to do something in a medically related field.
What I learned
Machine learning can be used to solve many classes of problems, and many machine learning problems can be made easier by combining feature values into a single vector.
What's next for Efficient ASD Detection using Machine Learning
Support for user inputs can be easily added next. They were not implemented in this version due to a lack of time. Using university resources in order to gain access to the medical datasets already available online would help to train the model better. Creating a user interface, or possibly a web application would be useful to collect data.
Log in or sign up for Devpost to join the conversation.