anhbiphan / project_chicago_traffic_crash

Chicago-Traffic-Crash-Project

Chicago Traffic-Crash Data Link:

https://data.cityofchicago.org/Transportation/Traffic-Crashes-Crashes/85ca-t3if

Project Details:

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.

Goals

• 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".

Problems:

• 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.

Solutions:

• Effective Googling for assistance.
• Experiementing with the code to check if they worked correctly.

Recommendations for further developments:

• 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.

Project Findings:

• 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.
  

Models Used:

• 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.

Getting Started

1. Clone this repo (for help see this tutorial).
2. Raw Data is being kept https://data.cityofchicago.org/Transportation/Traffic-Crashes-Crashes/85ca-t3if
3. CarCrashData.ipynb and Traffic_Crash-Anh.ipynb will contain data cleaning, findings, and visuals.
4. mytree.png will contain visual of DecisionTreeClassifier split parameters.

Project Members:

Name	Github

| |Anh Phan |Jesus Fuerte