基于Python/R的2018年1-5月限定条件海景房产均价预测技术问询

Got it, let's tackle this problem step by step. First, we need to clean up that unstructured "views" column to reliably flag all sea/beach/bay-facing properties. Then we'll filter down to the exact subset you need (200-300 sq.m., sea-view), and finally build a time-series forecast for Jan-May 2018.

Python 实现步骤 #

1. 数据清洗：识别海景房产 #

Those variant sea-view terms are a pain, but regex will save the day here. Just replace 'view_column' with your actual column name in the dataset:

import pandas as pd

# 加载你的数据集（替换成你的文件路径）
df = pd.read_csv('your_property_data.csv')

# 定义匹配所有海景变体的正则表达式（忽略大小写）
sea_view_pattern = r'(sea view|seaview|ocean view|water view|views to the sea|views of the sea|facing the sea|views to the beach|over the whole bay|over the bay)'

# 添加标记列：1表示海景房产，0表示非海景
df['has_sea_view'] = df['view_column'].str.contains(sea_view_pattern, case=False, na=False).astype(int)

2. 筛选目标数据集 #

Now narrow down to properties between 200-300 sq.m. and with sea views, then aggregate historical data into monthly averages (since we're forecasting monthly prices):

# 替换成你的建筑面积列名、日期列名和价格列名
df_filtered = df[
    (df['floor_area'] >= 200) & 
    (df['floor_area'] <= 300) & 
    (df['has_sea_view'] == 1)
].copy()

# 按月份聚合均价，适配时间序列模型格式
df_monthly = df_filtered.groupby(pd.Grouper(key='date', freq='M'))['price'].mean().reset_index()
df_monthly.columns = ['ds', 'y']  # Prophet要求的列名格式

3. 时间序列预测（用Prophet） #

Prophet is perfect for monthly time-series forecasts—it handles seasonality automatically. We'll predict the average price for Jan-May 2018:

from prophet import Prophet

# 初始化并拟合模型
model = Prophet(seasonality_mode='additive')
model.fit(df_monthly)

# 创建2018年1-5月的预测日期
future_dates = model.make_future_dataframe(periods=5, freq='M')
future_dates = future_dates[future_dates['ds'].dt.year == 2018].head(5)

# 生成预测结果
forecast = model.predict(future_dates)

# 提取最终预测的均价
predicted_avg_prices = forecast[['ds', 'yhat']]
print(predicted_avg_prices)

R 实现步骤 #

1. 数据清洗：识别海景房产 #

We'll use stringr for regex matching to catch all those sea-view variants. Replace view_column with your actual column name:

library(tidyverse)

# 加载数据集
df <- read_csv("your_property_data.csv")

# 定义海景正则模式（忽略大小写）
sea_view_pattern <- regex(
  "sea view|seaview|ocean view|water view|views to the sea|views of the sea|facing the sea|views to the beach|over the whole bay|over the bay",
  ignore_case = TRUE
)

# 添加标记列
df <- df %>% mutate(has_sea_view = as.integer(str_detect(view_column, sea_view_pattern)))

2. 筛选目标数据集 #

Filter for the 200-300 sq.m. range and sea views, then roll up historical data into monthly averages:

# 替换成你的建筑面积列名、日期列名和价格列名
df_filtered <- df %>%
  filter(floor_area >= 200, floor_area <= 300, has_sea_view == 1)

# 按月份聚合均价，适配Prophet格式
df_monthly <- df_filtered %>%
  mutate(month = floor_date(date, "month")) %>%
  group_by(month) %>%
  summarize(avg_price = mean(price, na.rm = TRUE)) %>%
  rename(ds = month, y = avg_price)

3. 时间序列预测（用Prophet） #

Use the R version of Prophet for consistent, easy-to-implement forecasts:

library(prophet)

# 初始化并拟合模型
model <- prophet(df_monthly, seasonality.mode = "additive")

# 创建2018年1-5月的预测日期
future_dates <- make_future_dataframe(model, periods = 5, freq = "month")
future_dates <- future_dates %>% filter(year(ds) == 2018) %>% slice(1:5)

# 生成预测结果
forecast <- predict(model, future_dates)

# 提取最终预测的均价
predicted_avg_prices <- forecast %>% select(ds, yhat)
print(predicted_avg_prices)

内容的提问来源于stack exchange，提问作者Bruno Benevides