Using Machine Learning to Detect Heart Attacks

Inspiration

This research was conducted to determine whether or not medical issues can be easily detected or prevented by machines with relatively small amounts of data. In this case, research was conducted using machine learning and a set of vitals at a point in time.

What it does

After training, the vitals were able to be fed into the algorithm, producing accurate results almost 80% of the time.

How we built it

Using the KNN training algorithm and a dataset consisting of various vitals and the result (1 for heart attack, 0 for nothing), the algorithm was trained with a "training set" (80% of the dataset) and a "test set" (20% of the dataset).

Challenges we ran into

This was my first time using machine algorithm completely on my own with no help other than what I learned during the CS1 course. I had some problems finding a dataset with a large amount of variables like I wanted, and setting up the custom test inputs took a while. However, I got it eventually. Additionally, I was supposed to be placed with teammates, but due to external problems, I was on my own, and I was only able to start late on July 18th, giving me about 3 days to finish the research paper and the algorithm itself. I was able to finish in time (barely) with help from David Chung.

Accomplishments that we're proud of

As I stated earlier, this was my first time ever making a machine learning algorithm without someone directly helping me with the program. David was did give very helpful advice, but the program itself was organized and written by me, completely on my own. I was expecting the results to be somewhere around the 50-60% range, but the fact that the data was so accurate surprised me. I am very satisfied with how the program came out, and I hope the judges are too!

What we learned

I learned more about a heart attack than I thought possible while researching what dataset would be most accurate and for the introduction section of my research paper. Fun fact, the average age for a heart attack is 7 years younger for males than females! I also learned how to actualy input data into the algorithm, instead of simply training it.

What's next for Using Machine Learning to Detect Heart Attacks

This simple algorithm could easily be integrated into other pieces of technology, such as smart watches, allowing for safe, quick, accurate, and convenient health monitoring. It is important to note that this is not limited to just heart attacks. This type of algorithm could easily be adapted as early warning systems for other conditions, permitting users to have their health monitored in several ways at all times with no effort on their part, letting more and more people stay healthy. The only possible limitation is how to get the proper vitals for detection, some of which aren’t as simple as a simple heart rate scan.

Built With

• kaggle
• kaggle/input/heart-attack-analysis-prediction-dataset/heart.csv
• python