COMP 309-403 Group Project #2

Bicycle Theft Prediction API & Frontend

An advanced machine learning API that predicts whether a stolen bicycle is likely to be recovered or remain stolen, based on historical bicycle theft data. View visualizations and predictions on the frontend against data. Enter your own bike data to see predictions.

Features

• Theft Prediction: Get probability scores for bicycle recovery using machine learning
• Location Services: Convert latitude/longitude to neighborhood information
• Field Options: Access all valid options for bike makes, models, colors, and more
• Data Preprocessing: Automatic handling of data formatting and encoding

API Endpoints

1. Prediction Endpoint (/predict)

Make predictions about bicycle theft outcomes:

POST /predict

Required fields:

• BIKE_TYPE: Type of bicycle (REGULAR, MOUNTAIN, or RACING)
• PREMISES_TYPE: Type of premises where incident occurred
• LOCATION_TYPE: Type of location
• OCC_DATE: Date of occurrence (YYYY-MM-DD)
• OCC_DOW: Day of week
• OCC_HOUR: Hour of occurrence (0-23)
• OCC_DOY: Day of year (1-366)
• REPORT_DATE: Date reported (YYYY-MM-DD)
• HOOD_140: Neighborhood code
• NEIGHBOURHOOD_140: Neighborhood name

Optional fields:

• BIKE_MAKE: Make of the bicycle
• BIKE_MODEL: Model of the bicycle
• BIKE_SPEED: Number of speeds
• BIKE_COLOUR: Color of the bicycle
• BIKE_COST: Cost of the bicycle

2. Neighborhood Lookup (/neighbourhood)

Get neighborhood information from coordinates:

POST /neighbourhood

Required fields:

• latitude: Latitude of the location
• longitude: Longitude of the location

3. Field Options (/options)

Get all valid options for input fields:

GET /options

4. Neighborhood Lookup (/api/thefts-over-time)

5. Neighborhood Lookup (/api/return-rate)

6. Neighborhood Lookup (/api/seasonal-analysis)

7. Neighborhood Lookup (/api/time-analysis)

8. Neighborhood Lookup (/api/value-analysis)

9. Neighborhood Lookup (/api/model-metrics)

##Webpage Endpoints

1. Home (/)

2. Prediction (/predict)

3. Model Dashboard (/model-performance)

4. About (/about)

Installation

1. Clone the repository:

git clone git@github.com:hodanysf/Group5-Project2.git

###Frontend

1. cd Frontend

2. npm install

3. npm start

4. Open http://localhost:3000 in your browser

###Backend

1. cd BicycleTheftPrediction

2. Create a virtual environment (recommended):

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

Requirements

• Python >= 3.8
• Dependencies:
  • numpy >= 1.21.0
  • pandas >= 1.3.0
  • scikit-learn >= 0.24.2
  • xgboost >= 1.4.2
  • imbalanced-learn >= 0.8.1
  • matplotlib >= 3.4.3
  • seaborn >= 0.11.2
  • flask >= 2.0.1
  • flask-restx >= 1.1.0
  • joblib >= 1.0.1
  • requests >= 2.26.0

Usage

1. Start the API server:

cd api/
python -m flask --app app.py run

2. Access the API documentation:

http://localhost:5000/docs

3. Make a prediction:

import requests

data = {
    "BIKE_TYPE": "REGULAR",
    "PREMISES_TYPE": "House",
    "LOCATION_TYPE": "Single Home",
    "OCC_DATE": "2023-01-01",
    "OCC_DOW": "Sunday",
    "OCC_HOUR": 14,
    "OCC_DOY": 1,
    "REPORT_DATE": "2023-01-02",
    "HOOD_140": "080",
    "NEIGHBOURHOOD_140": "Palmerston-Little Italy (80)"
}

response = requests.post("http://localhost:5000/predict", json=data)
print(response.json())

Group Members

• Hodan Ahmed Yusuf
• Dimitrios Avgerakis
• Chris Busse
• Gabriel Normand
• Nabanita Saha

ReadME based on original by Mindful-Developer