實價登錄房價分析¶

from google.colab import drive
drive.mount('/content/drive')

# Go to your google drive folder
os.chdir('/content/drive/MyDrive/Colab Notebooks/Intro_machine_learning/House_price_analysis') 
os.listdir()

# Crawl data
# Better store in your GDrive, so don't need to re-do it every time
def real_estate_crawler(year, season):
    if year > 1000:
        year -= 1911

    # download real estate zip file
    res = requests.get("https://plvr.land.moi.gov.tw//DownloadSeason?season="+str(year)+"S"+str(season)+"&type=zip&fileName=lvr_landcsv.zip")

    # save content to file
    fname = str(year)+str(season)+'.zip'
    open(fname, 'wb').write(res.content)

    # make additional folder for files to extract
    folder = 'real_estate' + str(year) + str(season)
    if not os.path.isdir(folder):
        os.mkdir(folder)

    # extract files to the folder
    with zipfile.ZipFile(fname, 'r') as zip_ref:
        zip_ref.extractall(folder)

    time.sleep(10)

for year in range(102, 110):
    for season in range(1,5):
        print('crawl ', year, 'Q', season)
        real_estate_crawler(year, season)

## 下載台北思源黑體並命名taipei_sans_tc_beta.ttf，移至指定路徑
!wget -O taipei_sans_tc_beta.ttf https://drive.google.com/uc?id=1eGAsTN1HBpJAkeVM57_C7ccp7hbgSz3_&export=download
!mv taipei_sans_tc_beta.ttf /usr/local/lib/python3.7/dist-packages/matplotlib//mpl-data/fonts/ttf

# title 數據分析
from matplotlib.font_manager import FontProperties
import matplotlib.pyplot as plt 

# 自定義字體變數
myfont = FontProperties(fname='/usr/local/lib/python3.7/dist-packages/matplotlib/mpl-data/fonts/ttf/taipei_sans_tc_beta.ttf')
myfont.set_size(15)

import os
import pandas as pd

# 歷年資料夾
dirs = [d for d in os.listdir() if d[:4] == 'real']

dfs = []

for d in dirs:
    df = pd.read_csv(os.path.join(d,locToLetter[location] + '_lvr_land_a.csv'), index_col=False)
    df['Q'] = d[-1]
    dfs.append(df.iloc[1:])
    
df = pd.concat(dfs, sort=True)

# 新增交易年份
df['year'] = pd.to_numeric(df['交易年月日'].str[:-4], errors='coerce') + 1911

# 平方公尺換成坪
df['單價元平方公尺'] = df['單價元平方公尺'].astype(float)
df['單價元坪'] = df['單價元平方公尺'] * 3.30579

# 建物型態
df['建物型態2'] = df['建物型態'].str.split('(').str[0]

# 刪除有備註之交易（多為親友交易、價格不正常之交易）
df = df[df['備註'].isnull()]

# 將index改成年月日
df.index = pd.to_datetime(df['year'].astype(str) + df['交易年月日'].str[-4:] ,errors='coerce')
df.sort_index(inplace=True)
df.head(2)

plt.rcParams['figure.figsize'] = (10, 6)

prices = {}
for district in set(df['鄉鎮市區']):
    cond = (
        (df['主要用途'] == '住家用')
        & (df['鄉鎮市區'] == district)
        & (df['單價元坪'] < df["單價元坪"].quantile(0.95))
        & (df['單價元坪'] > df["單價元坪"].quantile(0.05))
        )
    groups = df[cond]['year']
    prices[district] = df[cond]['單價元坪'].astype(float).groupby(groups).mean().loc[2012:]
    
price_history = pd.DataFrame(prices)
price_history.plot()
plt.title('各區平均單價',fontproperties=myfont)
plt.legend(prop=myfont)
plt.show()

district_price = df.reset_index()[['鄉鎮市區', '單價元坪']].dropna()
district_price = district_price[district_price['單價元坪'] < 2000000]

import plotly.express as px
fig = px.histogram(district_price, x="單價元坪", color="鄉鎮市區", marginal="rug", nbins=50) # can be `box`, `violin`)

# Overlay both histograms
fig.update_layout(barmode='overlay')
# Reduce opacity to see both histograms
fig.update_traces(opacity=0.75)
fig.show()

types = ['華廈', '公寓', '住宅大樓', '套房']#'店面', '套房', '其他', '農舍', '倉庫', '廠辦', '透天厝', '工廠']

building_type_prices = {}
for building_type in types:
    cond = (
        (df['主要用途'] == '住家用')
        & (df['單價元坪'] < df["單價元坪"].quantile(0.8))
        & (df['單價元坪'] > df["單價元坪"].quantile(0.2))
        & (df['建物型態2'] == building_type)
        )
    building_type_prices[building_type] = df[cond]['單價元坪'].groupby(df[cond]['year']).mean().loc[2012:]
pd.DataFrame(building_type_prices).plot()
plt.title('不同建物的價格',fontproperties=myfont)
plt.legend(prop=myfont)
plt.show()

import numpy as np
mean_value = price_history.mean()

gain = (price_history.iloc[-1] /  price_history.iloc[0])
mean = price_history.mean()

compare = pd.DataFrame({'上漲幅度': gain, '平均價格': mean}).dropna()
corr = (compare.corr().iloc[0,1])

print('相關性：', corr)
if corr > 0:
  print('意涵：價格越高越保值')
else:
  print('意涵：價格越低越保值')
print()
print('各區平均價格')
mean.sort_values()

gain = (price_history.iloc[-1] /  price_history.iloc[0])
mean = price_history.mean()

compare = pd.DataFrame({'上漲幅度': gain, '平均價格': mean}).dropna()
print(compare.corr().iloc[0,1])

compare