https://data.cityofchicago.org/Transportation/Traffic-Crashes-Crashes/85ca-t3if
Data Source: Based on Chicago Police Department Crash Dataset. Data Breakdown: Dataset describes traffic crash parameters, including date of crashes, street name, weather condition, posted speed limit, traffic way type, types of road defects etc.
- Create a model that predicted and classify if a crash is "Rear End" crash or not. Dataset parameters were used to predict and distinguish between types of crashes such as "Rear End".
- LabelEncoding, OneHotEncoding and merging dataframes to obtian a cummulative dataset that had everything we needed.
- Losing data after merging or adding columns.
- Large amounts of 'Unkown' and NaNs.
- Effective Googling for assistance.
- Experiementing with the code to check if they worked correctly.
- Grab Vehicle Dataset and Driver/Passenger Dataset. Combining them with our Traffic Crash dataset.
- Binning and Clustering data in the future to find location of crashes.
- Improve model.
-
The most important feature that played a big role in classifying was 'PRIM_CONTRIBUTORY_CAUSE'.
'UNABLE TO DETERMINE', 'FAILING TO YIELD RIGHT-OF-WAY', 'FOLLOWING TOO CLOSELY', 'NOT APPLICABLE', 'IMPROPER OVERTAKING/PASSING', 'IMPROPER BACKING', 'FAILING TO REDUCE SPEED TO AVOID CRASH', 'IMPROPER LANE USAGE', 'IMPROPER TURNING/NO SIGNAL', 'DRIVING SKILLS/KNOWLEDGE/EXPERIENCE', 'WEATHER', 'DISREGARDING TRAFFIC SIGNALS', etc.
- DecisionTreeClassifer : Based on 1 tree, we wanted to identify where the splits were being made.
- RandomForestClassifer : We used RandomForest to see if the model can become more accurate and better at predicting classes if it was given more trees to split on.
- Clone this repo (for help see this tutorial).
- Raw Data is being kept https://data.cityofchicago.org/Transportation/Traffic-Crashes-Crashes/85ca-t3if
- CarCrashData.ipynb and Traffic_Crash-Anh.ipynb will contain data cleaning, findings, and visuals.
- mytree.png will contain visual of DecisionTreeClassifier split parameters.
Name | Github |
---|
| |Anh Phan |Jesus Fuerte