Inspiration For Our Project
From watching an informative news segment on the issue of overprescription of opioids leading to high mortality rates, we were inspired to create a website that would create and spread awareness about the overwhelming opioid crisis. In the news segment, it was described that some surgeons wrote prescriptions for more than 100 pills in the week post-surgery, exceeding current guidelines that call for zero to ten pills. This overprescription frequently leads to opioid addiction, associated with a death rate of 28%. In 2016, deaths linked to opioid overuse numbered 42,249 in total. As a result, we developed OP.IO.ID, a website that would output the predicted lifespan based on the information entered into a form. By doing so, we hoped to draw awareness to the steps that patients can take to secure their own health regardless of the systematic issues that may pervade the healthcare facilities in their vicinity.
What We Learned
While attempting to transfer data across several programming languages and output mechanisms, we encountered challenges in processing and training models based on our data. As a principal component of our design, we aimed to develop a machine learning algorithm that uses RSM regression and implements TensorFlow. By combining PHP functionality to process user inputs and Python to analyze the data by back-propagating with RMS back-propagation to testing accuracy with MSE and MAE, we learned to test cases at each step of the process and consider the effects of each change in the context of the greater program. For instance, when converting data from PHP to Python, we had to consider the changes in syntax that followed, especially as we sought to train a model using pre-existing data and input test cases provided by the user. Scouring online forums for suggested techniques and collaborating with one another despite varying levels of experience, we learned to combine our varying skillsets in web development and machine learning to develop a viable product.
The Build Process
We developed a machine-learning algorithm using TensorFlow to predict the lifespan of an opioid-user based on demographics such as age, race, gender, location, and opioids currently in use. By training a model using a pre-existing dataset, we were able to predict the life-span of a user based on their personal characteristics, which they inputted into the front-end component of the website. The majority of the website was developed with HTML and CSS to create a straightforward user interface. We then used PHP to transfer data that would be used to generate the result listed on the results page.
Challenges We Faced
We encountered issues in melding the front-end and back-end components of the website. While we were able to build the machine learning algorithm and the front-end and back-end portions of the website, we struggled to actively transfer data from one side of the product to the other.