Boston_Housing_DataSet - MEDV predictor.

This repo contains files to code that deploys an API endpoint for a user to predict Median value of owner-occupied homes in $1000s (MEDV). The model was trained on the Boston Housing Dataset.

Build and retrain the model.

A notebook with steps about building, training and running the model is provided with steps shown.

Boston_Housing_Model_Building.ipynb

Please open the notebook shown above and run through the steps.

Directory Structure

The directory structure is provided below.

.
├── Boston_Housing_Model_Building.ipynb
├── Dockerfile
├── HousingData_test.csv
├── HousingData_train.csv
├── LICENSE
├── README.md
├── knn_imputer.joblib
├── requirements.txt
├── server.py
├── templates
│   ├── base.html
│   ├── index.html
│   └── output.html
└── xg_regressor.joblib

Build steps:

Listed below are different ways to build and run the system. There are a total of 3 ways to build and run the system. They are:

1. Build from the git repo.
2. Build using docker image.

---

1. Build from the git repo.

• Clone the project Git repo.

  foo@bar:~$ git clone https://github.com/likith11/Boston-Housing-DataSet-Predictor.git

• Move into the project directory.

  foo@bar:~$ cd Boston-Housing-DataSet-Predictor

• Create a virtual environment and activate it.

  foo@bar:~$ python3 -m venv
  foo@bar:~$ source venv/bin/activate

• Install requirements from requirements.txt

  (venv) foo@bar:~$ pip install -r requirements.txt

   

NOTE : Steps to run shown below in Steps To RUN the web server in To run locally using py file section:

---

 

2.) Pull from Docker Hub / Build Docker Image Locally:

2.1) Build using docker image.

 

• Pull image from docker hub.

  foo@bar:~$ docker pull likithponnanna/boston-housing-predictor

• Run the image that was pulled. Below port 5000 is mapped between the external and docker container.

  foo@bar:~$ docker run -p 5000:5000 boston-housing-predictor

NOTE : Steps to run also shown below in Steps To RUN the web server in To run the docker image section.

   

2.2 Build from docker hub locally using the DockerFile provided:

• To build the docker image using the dockerfile locally run the following command before initiating docker run.

  foo@bar:~$ git clone https://github.com/likith11/Boston-Housing-DataSet-Predictor.git
  foo@bar:~$ cd Boston-Housing-DataSet-Predictor
  foo@bar:~$ docker build -t boston-housing-predictor .

• Run the docker image with port specified.

  foo@bar:~$ docker run -p 5000:5000 boston-housing-predictor

NOTE : Steps to run also shown below in Steps To RUN the web server in To run the docker image section.

---

   

Steps To RUN the web server:

1. To run locally.
2. To run the docker image.

1.) To run locally using py file.

(venv) foo@bar:~$ python server.py

2.) To run the docker image.

Run the docker image with port specified.

   foo@bar:~$ docker run -p 5000:5000 boston-housing-predictor

Navigate through the web deployment:

There are 2 ways of using the web deployment.

1. Using GUI
2. Using API call

1.) Using GUI:

• Navigate to http://0.0.0.0:5000/ or http://localhost:5000/ (Replace port number if its a different port on your machine. The default port above is 5000).
• Two text fields with pre-filled test data along with a blue Process button is displayed.
• Change the input field to the required test input.
• Click on _Process__
• A new screen with Translucent alert box is shown with the corresponding MEDV predicted.
• Click on Retry? button to go back to the main screen and repeat the process.

2.)Using API call:

• API end point is deployed at http://0.0.0.0:5000/api/predict where POST request is accepted.
• The API accepts text inputs in the JSON format shown below

{
  "CRIM": 9.39063,
  "ZN": 0.0,
  "INDUS": 18.1,
  "CHAS": 0,
  "NOX": 0.74,
  "RM": 5.627,
  "AGE": 93.9,
  "DIS": 1.8172,
  "RAD": 24.0,
  "TAX": 666.0,
  "PTRATIO": 20.2,
  "B": 396.9,
  "LSTAT": 22.88
}

• An example curl command is

curl -d '{ "CRIM"  : 9.39063,  "ZN"    : 0.0,  "INDUS" : 18.1, "CHAS"  : 0, "NOX"    : 0.740,   "RM"    : 5.627,  "AGE"   : 93.9, "DIS"   : 1.8172, "RAD"   : 24.0, "TAX"  : 666.0, "PTRATIO":20.2, "B"     : 396.90, "LSTAT" : 22.880 }' -H 'Content-Type: application/json' http://0.0.0.0:5000/api/predict

• An API response json is returned with the the predicted MEDV value.

 

---