911 calls Forecast System (web app)
During my Insight Data Science project I developed an application to improve emergency response by forecasting 911 calls (911forecast.site). In particular I focused on medical emergency and used a dataset of 911 calls streamed in real-time from Montogmery Country, PA (http://montcoalert.org/gettingdata/).
The time-series machine learning model was developed in python by using sklearn toolkit. A new GEO-Forecasting machine learning model is under development (see news below)
System architecture
Data layer
This class develops the DB abstraction layer
Maps layer
This class was developped to determine the spatial distribution of 911 calls and the Geo forecast probability distribution (new feature coming soon)
Model-training layer
This module contain the code that implements and trains the time-series forecast-model. GradientBoosting and linear Regression were used. A new class containing the GEO-Forcasting for this layer is under implementation
Model-forecast layer
This module contain the code that implements the hour and day forecasting
911forecast web app
The web app was developped to create a first functional version of the 911 forecast system
[news]
A GEO-Forecasting model is under implementation
