Machine Learning
In order to make change at Penn, we started to create models to predict the Delaware River's pH levels. Through our previous experience and research, we decided on a N-HiTs model. Below is the actual vs. predicted graph for the last week of the Delware River's pH level. You can see that the model pretty accurately captures the trend in the pH level, with an mean squared error of around 0.035!
Predicted Values
.png)
Actual Values
The next steps are to refine the machine learning model and the data inputted in order to create even more accurate results. From there, we will try to analyze a longer horizon of pH values, for example two weeks or a month. We can then see whether there are any spikes or major changes in the pH values that may impact the biodiversity and even humans who use the Delaware River!
Wharton Data Science Academy
At the Wharton Data Science Academy, Abhay led a team to conduct a statistical analysis on the Delware River. Check it out below!
Wharton Data Science Academy Presentation
Data Models
Below there are data models that we generated using R Programming. The models show a time series of the pH level of that specific river, with the data being smoothed out for easier visuals. You can see the cyclical trend that a lot of the data demonstrates, and many of the data sets also show a negative trend. The pictures fade into new ones every 5 seconds, so be sure to stick around and see them all!
