Inspiration

This was one of the listed challenges. Although, when we were interacting with the sponsors, this project caught our eye and made us already think at that moment how are we gonna implement this.

What it does

We built a web app using which one can enter patient's data while making an appointment for him/her. and it will predict based on the patient's attribute, whether the person is going to actually show up or not for the appointment.

How we built it

We used kinds of tool to complete this end to end. We used PySpark to run machine learning algorithms on the input data. We picked Logistic regression technique to do this task as we analyzed that this is gonna be best suited one as per the problem scenario. We also used google-analytics api and google visualization api to analyze the data and easliy visualize it by making different kind of charts. We also designed a single page website, which takes up the input data to book an appointment and to predict at real time, if the person is gonna show up or not. To integrate all these modules in one place, we use Flask .

Challenges we ran into

While designing and developing out prediction engine, we ran into challenges like which features to take into consideration and how are they gonna impact our engine.Integrating all the different pieces was the also a challenging part.

Accomplishments that we're proud of

That our web app predicts in real time effectively and satisfies the problem use case effectively.

What we learned

A lot of things, as we ran into many challenges to implement it in such short period. Designing and implementing the prediction engine. Flask was certainly the major tool , that we used and learnt.

What's next for Northwell Predictive Analysis

The product that we aimed for is not just useful for the given use case, but it can be incorporated in many use case. Apart from the various similar appointment like use cases, the prediction engine can be tweaked and used in various predictive analysis. We also want to improve our prediction engine by using neural networks and other algorithms and see how it behaves. As far as this specific use case is considered, many other useful attributes can also be included. We can use the Google maps API and improve our predication accuracy considering the expected traffic scenario.

Share this project:

Updates