基于SVR与VE-PCA的住宅价值评估模型研究

庞瑜韬

 摘 要

“住”是人类生活的基本需求之一，随着我国经济社会的高速发展，公众生活品质不断提升，对居住条件的需求也随之提高，房地产企业为了在激烈的市场竞争中夺得一席之地，需要从以往的高周转模式转向满足人们需求的高价值模式。价值源于需求，房地产企业要开发出高价值产品，就必须充分掌握潜在客户对住宅的需求，但以往收集需求的常用方法难以准确定位潜在客户、难以大范围收集需求信息，使得房地产企业向高价值模式的转型面临困难。本文为了解决这一问题，通过挖掘用户在互联网上的行为数据、分析用户需求，以此建立住宅价值评估模型，为房地产企业开发高价值住宅项目提供建议。

本文抓取房地产网络平台在武汉地区新建楼盘的用户评论数据和评分数据，以评论数据构建住宅价格预测模型，以评分数据构建住宅性能价值模型，两者共同构成住宅价值评估模型。主要的研究内容包括两部分：（1）建立基于需求的住宅价格预测模型，该部分以用户评论数据为对象，以词频统计分析、Spearman相关分析等方法分析和提取需求，以SVR进行数据建模，研究结果表明SVR模型预测性能较好；（2）基于用户评分数据构建住宅价值模型，该部分以用户对住宅性能评分数据为对象，以VE-PCA进行数据分析，最终得到各新建楼盘的住宅性能价值。以价值工程为评估的理论基础，预测的住宅价格p与住宅性能价值v之比为住宅价值w，代表住宅产品满足用户需求程度与用户愿为其付出的价格之比，在项目开发前期，企业可以此作为一个指标进行方案比选，做到满足用户需求与其价格之间的平衡。

关键字：用户需求；房价预测；价值工程；支持向量机；主成分分析

Abstract

"Live" is one of the basic needs of human life, with the high-speed development of China's economic and social, improve the public quality of life, the living conditions of demand increases, real estate enterprises in order to win a place in the fierce market competition, from the previous high turnover of models to meet the demand of people's high value. Value originates from demand. To develop high-value products, real estate enterprises must fully grasp the housing demand of potential customers. However, the common methods of collecting demand in the past are difficult to accurately locate potential customers and collect demand information in a wide range, which makes the transformation of real estate enterprises to high-value model face difficulties. In order to solve this problem, this paper builds a residential value evaluation model by mining users' behavior data on the Internet and analyzing users' needs, so as to provide Suggestions for real estate enterprises to develop high-value residential projects.

This paper grabs the user comment data and score data of new buildings built on the real estate network platform in Wuhan area, constructs the housing price prediction model based on the comment data, and constructs the housing performance value model based on the score data, which together constitute the housing value assessment model. The main research contents include two parts :(1) establishing a demand-based housing price prediction model. This part takes user comment data as the object, analyzes and extracts demand by word frequency statistical analysis, Spearman correlation analysis and other methods, and conducts data modeling with SVR. The research results show that the SVR model has good predictive performance; (2) The residential value model is built based on the user score data. In this part, the residential performance score data is taken as the object, and ve-PCA is used for data analysis to finally obtain the residential performance value of each new building. On the basis of the theory of value engineering to evaluate and predict the housing price p and the ratio of the residential property value v for house value w, on behalf of residential products to meet user demand degree and the ratio of the user is willing to pay for the price, in the early stage of the project development, the enterprise can be as an indicator for scheme comparison, the balance between satisfying user demand and its price.

Key words: user demand; housing price forecast; value engineering; support vector machine; Principal Component Analysis.