This folder contains documents for capstone1 project of the Data Science Career Track at Springboard.
The objective was to predict the severity of the road accident based on various weather conditions, road amenities(e.g. stop signs, speed-bumps, junctions etc.), weekday, hour of the day, side of the road, state and geographic location.
Severity here refers to the effect of the accident on traffic flow: low severity accidents have little effect while high severity accidents result in higher obstructin of flow.
Upon comparison of four different types of classification models (SGD, KNN, LogisticRegression and Random Forest), the latter two were chosen for further analysis. A final 78% accuracy was achieved using Random Forest. The top four most important features were: longitude, latitude, pressure and windchill.
The image files are stored in a seperate folder, and the raw dataset can be found in Kaggle under US Accidents (3.5 million records)!.
List of documents, in chronological order:
- Capstone 1 Project Proposal
- Data Wrangling
- Wind Chill prediction tests
- Data Storytelling
- US Accidents Statistical Analysis
- US Accidents Modeling and InDepth Analysis
- Capstone 1 Milestone Report
- Capstone 1 Final Report
Tableau Public dashboard of Traffic Accidents.