Parking Citations Data Analysis

Sections

The repository contains the followings:

1. run.sh: A shell script to:
   1. Download the dataset from S3
   2. Install all Python requirements
   3. Train the data by calling train.py
   4. Launch model server by calling api.py & model_core.py
2. API Calls methods:
   1. api_client.py: a python script to call the model server
   2. curl example
   3. Postman collection
3. Notebook files
   1. model.ipynb: A notebook that contains process to find the best training approach and model evaluation
   2. data_analysis.ipynb: A notebook that contains EDA on raw data
   3. part_two.ipynb: A notebook that contains analysis related to the part two of the assessment
4. Dockerfile: To build a linux image that runs the model (optional)

How to run:

Clone the repo

git clone parking_citations_analysis.git

Option 1: Run locally:

Start a Python virtual env (optional)

In order to isolate your local python development environment for this project, you can take advantage of Python virtual environments virtual environments:

> pip install virtualenv
> cd parking_citations_analysis/
> virtualenv venv
> source venv/bin/activate

You can leave the virtual environment in your termial by:

> deactivate

Launch the server

The following shell script should download the required datasets (may take a few minutes), and train the data:

> ./run.sh

Important note: If you do not have the dataset already downloaded in both forms of CSV, and SQLite DB, you need to set the following configuration flags to True in the script:

#Should the dataset be downloaded?
download_data_enabled=1;

#Should the sqlite database be created?
create_sqlite_db_enabled=1;

If you want to re-train the model from the scratch, set the following configuration flat to True (Please note that trainign the model may take hours.) :

#Should the model be trained? 
train_model_enabled=1;

If you get the following error when running the script:

./run.sh: Permission denied

Add read/write/execution permissions to your file to everyone:

chmod 777 run.sh

The script will launch a local Flask server for the model.

If script runs with no issues, Flask should be running:

INFO:werkzeug: * Running on http://0.0.0.0:5000/ (Press CTRL+C to quit)

Option 2: Docker

If you have Docker installed, build the docker image using the docker file:

> cd parking_citations_analysis/
> docker build . -t citations

This takes a few minutes as the docker build process also downloads the dataset from S3 and creates a SQLite database copy out of it as well. Then you should be able to run the built container

> docker run -it -p 5000:5000 citations 

Call the model server with input data

For the pretrained model we only need to pass 4 columns: "Color", "Body Style", "Fine amount", "Plate Expiry Date" You have three options to call the model server:

Option 1: A Python script (easiest)

There is also a quick python script (api_client.py) that uses Python's http.client library to call the API. You can modify the payload in code and run the script to call the model server.

Option 2: CURL

Send your sample data from a separate terminal using curl:

curl --header "Content-Type: application/json" \
     --request POST \
     --data '{"Color": "WH", "Body Style": "PA", "Fine amount": 50.0,"Plate Expiry Date": 200304.0}' \
     http://localhost:5000/model

You should be able to see the response in form of a JSON object:

{"popular_make_probability":55.23}

Option 3: Postman

If you have Postman installed, you can import the sample Postman collection in the repository where the API calls are pre-defined:

Parking-Citatations-Postman-Call.postman_collection.json