Spaces:

sivakornchong
/

HDB_resale_predict

Running

App Files Files Community

HDB_resale_predict / main.py

sivakornchong

add code to calculate df

7c9d88c 6 months ago

raw history blame contribute delete

No virus

4.18 kB

	import json
	import requests
	from misc import nearest_mrt
	import pickle
	import os
	import pandas as pd
	import datetime
	from datetime import datetime

	def findlast10(postal):
	df = pd.read_json("data/data_features.json", lines=True)
	df_filtered = df[df['Postal']==str(postal)]

	df_output=df_filtered.sort_values(by='transaction_yr', ascending=False).head(10).reset_index(drop=True)

	storey_dict = {
	'01 TO 03': 1, '04 TO 06': 2, '07 TO 09': 3, '10 TO 12': 4,
	'13 TO 15': 5, '16 TO 18': 6, '19 TO 21': 7, '22 TO 24': 8,
	'25 TO 27': 9, '28 TO 30': 10, '31 TO 33': 11, '34 TO 36': 12,
	'37 TO 39': 13, '40 TO 42': 14, '43 TO 45': 15, '46 TO 48': 16,
	'49 TO 51': 17
	}

	# Swap keys and values using dictionary comprehension
	swapped_dict = {value: key for key, value in storey_dict.items()}

	df_output['storey_height']=df_output['storey_height'].apply(lambda x: swapped_dict[x])
	df_out = df_output[['transaction','area','storey_height','resale_price']]

	return df_out


	def main_fn(Postal_,age_,town_,storey_,room_):
	#Load model
	filename = 'finalized_model2.sav'
	if os.path.exists("./finalized_model2.sav"):
	model = pickle.load(open(filename, 'rb'))
	print('loaded model')
	else:
	print('failed loading model')

	#extract feature names
	feature_names = model.feature_names_in_.tolist()
	input = [0]*len(feature_names)

	#Set up mrt_list
	mrt_name = []
	mrt_loc = []
	with open('data/mrt_list.json', 'r') as file:
	for line in file:
	item = json.loads(line)
	mrt_name.append(item['MRT'])
	loc = tuple([float(i) for i in item['location']])
	mrt_loc.append(loc)

	# #Test input
	# Postal_,age_,town_,storey_,room_ = 680705, 30, 'CHOA CHU KANG', 12, '5 ROOM'

	##POSTAL
	Postal_input = int(Postal_)
	# Postal_input = 680705
	input[feature_names.index('Postal')] = Postal_input

	##DISTANCE TO MRT
	search_term = Postal_input
	query_string= 'https://www.onemap.gov.sg/api/common/elastic/search?searchVal={}&returnGeom=Y&getAddrDetails=Y&pageNum=1'.format(search_term)
	resp = requests.get(query_string)
	data = json.loads(resp.content)
	print(query_string)
	print(data)
	chosen_result = data['results'][0]

	#Calculate the distance to nearest MRT
	distance_km, nearest_mr = nearest_mrt(chosen_result['LATITUDE'], chosen_result['LONGITUDE'], mrt_name, mrt_loc)
	input[feature_names.index('distance_mrt')] = distance_km

	##STOREY
	#Height is input, but then converted to the scale we used for iterating model
	height_input = int(storey_)
	# height_input = 51
	Height = (height_input+2)//3
	input[feature_names.index('storey_height')] = Height

	##Town
	input[feature_names.index("town")]=town_

	##Room
	input[feature_names.index("flat_num")]=room_

	##AGE/ TRANSACTION YEAR [Current default to 2024]
	age_input = int(age_)
	# age_input = 30

	# Get the current date
	current_date = datetime.now()

	input[feature_names.index('age_transation')] = age_input
	input[feature_names.index('transaction_yr')] = current_date.year #Default to 2024 first

	# Create final_dataframe as input to model
	Actual = dict(zip(feature_names,input))
	Actual_df = pd.DataFrame(Actual, index=[0])

	# Use model to predict adjusted price
	resale_adj_price = model.predict(Actual_df)[0]

	# Readjust back to actual price
	# Calculate the quarter
	quarter = (current_date.month - 1) // 3 + 1
	# Format the quarter in the desired format
	formatted_quarter = f"{quarter}Q{current_date.year}"
	RPI_pd = pd.read_csv('data/RPI_dict.csv', header=None)
	RPI_dict = dict(zip(RPI_pd[0], RPI_pd[1]))
	RPI = float(RPI_dict[formatted_quarter])
	price = resale_adj_price*(RPI/133.9)

	# Obtain the last 10 transactions with the same postal address
	df = findlast10(Postal_input)

	return (int(price), df)

	if __name__ == "__main__":
	Postal_,age_,town_,storey_,room_ = 680705, 30, 'CHOA CHU KANG', 12, '5 ROOM'
	item = main_fn(Postal_,age_,town_,storey_,room_)
	print(item)