Skip to content

Latest commit

 

History

History
110 lines (89 loc) · 4.02 KB

README.md

File metadata and controls

110 lines (89 loc) · 4.02 KB
Raw

Canada house

The purpose in this project is digging out raw data of selling houses from this site. Then insert them into our local database and visualize them into our django-base website. And finally we are going to implement a learner agent that estimates price of houses by their features. By the way this project is going to be developed only for fun and learning purpose. Thanks to dear Jadi for this idea.

How to run

To run project in development mode, Just use steps below:

  1. Install python3, pip3, virtualenv in your system.
  2. Make development environment ready using commands below;
git clone https://github.com/sajadkhosravani1/canadaHouse && cd canadaHouse
virtualenv -p python3 venv # Create virtualenv named venv
source venv/bin/activate
pip install -r requirements.txt
python manage.py migrate
  1. Fetch data from https://gopalsharma.ca by following command:
python manage.py fetch 2000
  1. Train the learner agent by following command
python manage.py train 80
  1. Run the project using python manage.py runserver
  2. Go to http://localhost:8000 to see the home page.

Digging out data

The purpose data are observable at here:https://gopalsharma.ca/1000000-1500000. You can also see the hyper texts in your browser in inspect mode. But you would not be able to receive the relative hyper texts by code simply with a request to the url. And you will see the container of our purpose data with no content. (The container's tag-id is carmen which would be empty). So it's obvious that there's some js that downloads the content in the browser, as it arrives to the client side. So we have to find that js code to see where it fetches data from. I have implemented a django command that finds all of imported js files' urls. You can simply use it this way:

python manage.py get-js-links

After taking a look at the js files or monitoring the calling urls from your browser, you will find the api that exempts us from parsing html tags.

Fetching records

I have implemented a manage.py command that fetches data from API and stores them into tables.

command source: /gopal/management/commands/fetch.py

You can simply use it as following.

python manage.py fetch <limit>

Use an integer instead <limit> so api can limit fetched objects count. We can make some configurations in fetch.py file.

lazy mode

In lazy mode we would assume that when we bump into a case that we have already saved that means we have already saved the rest, so breaks the loops. because the cases are sorted by date. But when lazy mode is off, every single objects will be checked. You can simply turn on/off lazy mode by changing the lazy variable to True/False in fetch.py file.

Price limit

We can also config the min and max price, to limit and speed-up fetch operation, as simple as editing the priceRange list variable, in fetch.py file. The first item is Min value and the Second item is Max value in dollars.

delete_before command

We can delete out-dated house records by the following command.

python manage.py "<datetime>"

<datetime> is Date and time with YYYY-MM-DD hh:mm:ss format or only date in YYYY-MM-DD format.

Train

We can simply train the agent to estimate the house price by train command. The train result will be saved in houses_trained.pkl file. The result will be read and used when user sends estimate request to houses/estimate web page. We can use train the agent by following command:

python manage.py train <percentage>

<percentage> is an integer specifying train set percentage ratio than the all of dataset.

When the train operation finished the r2 score of test-set estimating will be reported.

TODO list

  • Digging out data
  • Data base modeling
  • Fetch command
  • Delete command
  • Train command
  • URL dispatching
  • Houses list view and template
  • Houses estimate view and template
  • Home view