StatisticalBaseModel.py

from typing import Tuple, TypedDict, Optional
import datetime


class ProcessedSynapse(TypedDict):
    id: Optional[str]
    nextplace_id: Optional[str]
    property_id: Optional[str]
    listing_id: Optional[str]
    address: Optional[str]
    city: Optional[str]
    state: Optional[str]
    zip_code: Optional[str]
    price: Optional[float]
    beds: Optional[int]
    baths: Optional[float]
    sqft: Optional[int]
    lot_size: Optional[int]
    year_built: Optional[int]
    days_on_market: Optional[int]
    latitude: Optional[float]
    longitude: Optional[float]
    property_type: Optional[str]
    last_sale_date: Optional[str]
    hoa_dues: Optional[float]
    query_date: Optional[str]
    market: Optional[str]


class StatisticalBaseModel:

    def __init__(self):
        self._load_model()

    def _load_model(self):
        """
        Perform any actions needed to load the model.
        EX: Establish API connections, download an ML model for inference, etc...
        """
        print("Loading model...")
        # Optional model loading
        print("Model loaded.")

    def _get_average_for_market(self, market: str) -> int:
        """
        Get the average days on market for a house in a given market
        :param market: the housing market
        :return: the average days on market
        """
        # You probably want to update this based on the current season. Houses sell faster in the summer.
        # Add more logic for other housing markets!
        if market == 'San Francisco':
            return 23
        elif market == 'Los Angeles':
            return 68
        elif market == 'Seattle':
            return 27
        elif market == 'Austin':
            return 78
        elif market == 'Houston':
            return 73
        elif market == 'Chicago':
            return 25
        elif market == 'New York':
            return 20
        elif market == 'Denver':
            return 24
        return 34


    def _sale_date_predictor(self, input_data: ProcessedSynapse):
        """
        Calculate the expected sale date based on the national average
        :param days_on_market: number of days this house has been on the market
        :return: the predicted sale date, based on the national average of 34 days
        """
        if 'days_on_market' not in input_data:
            return datetime.date.today() + datetime.timedelta(days=1)

        if 'market' not in input_data:
            average = 34

        else:
            average = self._get_average_for_market(input_data['market'])

        days_on_market = input_data['days_on_market']
        if days_on_market < average:
            days_until_sale = average - days_on_market
            sale_date = datetime.date.today() + datetime.timedelta(days=days_until_sale)
            return sale_date
        else:
            return datetime.date.today() + datetime.timedelta(days=1)

    def _get_price_multiplier(self, market: str) -> float:
        """
        Calculate the price multiplier based on the market
        :param market: the marked the house is in
        :return: the multiplier for the predicted price
        """
        # You may want to add more logic to check zipcode for more precise price multipliers
        # Add more logic for other housing markets!
        if market == 'San Francisco':
            return 1.18  # 18% above listing
        elif market == 'Los Angeles':
            return 1.2  # 22% above listing
        elif market == 'Seattle':
            return 1.13  # 13% above listing
        elif market == 'Austin':
            return 1.11  # 11% above listing
        elif market == 'Houston':
            return 1.15  # 15% above listing
        elif market == 'Chicago':
            return 1.12  # 12% above listing
        elif market == 'New York':
            return 1.05  # 5% above listing
        elif market == 'Denver':
            return 1.11  # 11% above listing
        return 1.0

    def run_inference(self, input_data: ProcessedSynapse) -> Tuple[float, str]:
        """
        Predict the sale price and sale date for the house represented by `input_data`
        :param input_data: a formatted Synapse from the validator, representing a currently listed house
        :return: the predicted sale price and predicted sale date for this home
        """
        listing_price = float(input_data['price']) if 'price' in input_data else 1.0
        sale_multiplier = self._get_price_multiplier(input_data['market']) if 'market' in input_data else 1.0
        predicted_sale_price = listing_price * sale_multiplier
        predicted_sale_date = self._sale_date_predictor(input_data)
        predicted_sale_date = predicted_sale_date.strftime("%Y-%m-%d")
        return predicted_sale_price, predicted_sale_date