Property price prediction API

Table of Contents

1. About
2. Setup
3. Label Encoded Values
4. Endpoints

About

This repo contains files and Jupyter notebooks for the price prediction of HDB resale flats in Singapore.

The data is obtained from data.gov.sg (Singapore publicly available data) and we are using a Flask API for price prediction, price trend calculation and checking for data updates.

---

The repo is organized as follows:

• /data: All saved data for model training is stored here.

• /templates: HTML files for the webpage and form.

• /training_notebooks:

  1. 1_data_extraction.ipynb: Jupyter notebook for data extraction

     1. This will help us obtain SavedData.csv needed for the second notebook)
     2. This notebook can be used in place of the /checkupdates endpoint if desired.

  2. 2_data_preprocessing.ipynb

     1. Jupyter notebook for data preprocessing and model training

• /utils: Contains helper and prediction functions for the various endpoints.

---

2 ways you can view the prediction:

1. Entering values and submitting the form on the main page.
2. Send in a POST request. See Endpoints for sample input and outputs.

Setup

1. Create the virtual environment:

Mac:

python3 -m venv env

Windows:

py -m venv env

python -m venv <env name>

2. Activate the virtual environment

Mac:

source env/bin/activate

Windows:

<env path>\Scripts\activate

3. Install required packages:

Mac:

python3 -m pip install -r requirements.txt

Windows:

py -m pip install -r requirements.txt

4. Run Flask app in debugger mode:

Mac:

python3 app.py

Windows:

python app.py 

5. IMPORTANT Run all cells of the 2_data_preprocessing.ipynb notebook to obtain the model.pkl and month_encoder.pkl files. These files are required for the price prediction model.

Endpoints

By default, the app is hosted on http://127.0.0.1:5000/. Loading this URL on a web browser will return a sample form that you can use to test and view the predicted prices (in HTML format).

1. /predict (REST POST):

• REST API endpoint for obtaining predicted price result (JSON). Request body values correspond to label encoded values. See Label Encoded Values for the encoded values.

Sample request body:

{
    "flattype": "5", 
    "flatmodel": "7", 
    "floorarea": "152", 
    "lease": "1992", 
    "storey": "1", 
    "town": "21",
    "startdate": "2018-07-01",
    "months": "25"
}

This will return an array of objects. The number of objects corresponds to number of months + 1, where the first object corresponds to the start date.

Sample response:

[
    {
        "date": 1530403200000,
        "predicted_price": 603055.9
    },
    {
        "date": 1533081600000,
        "predicted_price": 604578.61
    },
    {
        "date": 1535760000000,
        "predicted_price": 606101.31
    },
    {
        "date": 1538352000000,
        "predicted_price": 607624.02
    }
]

2. /pricetrend (REST GET)

• API endpoint for obtaining average price per sq m of HDB resale flats over the previous months.

Sample response:

[
    {
        "date": 1483228800000,
        "price_per_sq_m": 4523.77
    },
    {
        "date": 1485907200000,
        "price_per_sq_m": 4585.07
    },
    {
        "date": 1488326400000,
        "price_per_sq_m": 4624.39
    }
]

3. /checkupdates (REST GET)

• API endpoint for checking if there are new updates to the api on data.gov.sg, and if existing data has been successfully overwritten.

Sample response:

{
    "success": true,
    "updates": true
}

Label encoded values

We use label encoding for the request body inputs: town, flat_model, flat_type, storey. The mappings are encoded as follows:

Town:

{'ANG MO KIO': 0, 
  'BEDOK': 1, 
  'BISHAN': 2, 
  'BUKIT BATOK': 3, 
  'BUKIT MERAH': 4, 
  'BUKIT PANJANG': 5, 
  'BUKIT TIMAH': 6, 
  'CENTRAL AREA': 7, 
  'CHOA CHU KANG': 8, 
  'CLEMENTI': 9, 
  'GEYLANG': 10, 
  'HOUGANG': 11, 
  'JURONG EAST': 12, 
  'JURONG WEST': 13, 
  'KALLANG/WHAMPOA': 14, 
  'MARINE PARADE': 15, 
  'PASIR RIS': 16, 
  'PUNGGOL': 17, 
  'QUEENSTOWN': 18, 
  'SEMBAWANG': 19, 
  'SENGKANG': 20, 
  'SERANGOON': 21, 
  'TAMPINES': 22, 
  'TOA PAYOH': 23, 
  'WOODLANDS': 24, 
  'YISHUN': 25}

Flat model:

{'2-room': 0, 
  '3Gen': 1, 
  'Adjoined flat': 2, 
  'Apartment': 3, 
  'DBSS': 4, 
  'Improved': 5, 
  'Improved-Maisonette': 6, 
  'Maisonette': 7, 
  'Model A': 8, 
  'Model A-Maisonette': 9, 
  'Model A2': 10, 
  'Multi Generation': 11, 
  'New Generation': 12, 
  'Premium Apartment': 13, 
  'Premium Apartment Loft': 14, 
  'Premium Maisonette': 15, 
  'Simplified': 16, 
  'Standard': 17, 
  'Terrace': 18, 
  'Type S1': 19, 
  'Type S2': 20}

Flat type:

{'1 ROOM': 0, 
  '2 ROOM': 1, 
  '3 ROOM': 2, 
  '4 ROOM': 3, 
  '5 ROOM': 4, 
  'EXECUTIVE': 5, 
  'MULTI-GENERATION': 6}

Storey:

{'01 TO 03': 0, 
  '04 TO 06': 1, 
  '07 TO 09': 2, 
  '10 TO 12': 3, 
  '13 TO 15': 4, 
  '16 TO 18': 5, 
  '19 TO 21': 6, 
  '22 TO 24': 7, 
  '25 TO 27': 8, 
  '28 TO 30': 9, 
  '31 TO 33': 10, 
  '34 TO 36': 11, 
  '37 TO 39': 12, 
  '40 TO 42': 13, 
  '43 TO 45': 14, 
  '46 TO 48': 15, 
  '49 TO 51': 16}