融合模拟退火的随机森林房价评估算法

传统的随机森林房价评估算法存在着大量参数组合计算问题，参数的优劣对算法准确度影响很大。针对此问题，结合随机森林和模拟退火算法提出一种融合模拟退火的随机森林房价评估算法。首先，通过10次10折交叉验证法对参数进行敏感性测试，选择出对随机森林算法敏感的参数；然后结合模拟退火算法对敏感的参数迭代寻优，通过与网格搜索算法、随机搜索算法进行对比分析发现，在参数组合计算过程中，模拟退火算法在运行时间和算法准确率方面更优，弥补了网格搜索算法耗时过长和随机搜索算法低准确率的缺陷；最后，将融合模拟退火的随机森林算法应用于房价评估问题，构成新的房价评估算法。将新算法与传统随机森林房价评估算法进行了对比实验分析，结果表明，融合模拟退火的随机森林房价评估算法误差值减少，拟合优度值增加，评估的准确度得到了显著提升。

Housing prices evaluation using random forest algorithm combing with simulated annealing

The traditional housing prices evaluation which is using random forest algorithm has a large number of parameter selection problems and the parameters has great influence on the accuracy of the algorithm. In order to solve the above problem, this paper proposed a new algorithm combined the random forest algorithm and simulated annealing algorithm about the housing prices evaluation. Firstly, this paper tested the sensitivity of parameters by 10 times 10-cross-validation method and selected the parameters which were sensitive of the random forest algorithm. Secondly, it used the simulated annealing algorithm to optimize the sensitive parameters iteratively. Through comparing to the grid search algorithm and random search algorithm, it found the simulated annealing algorithm was better in the running time and algorithm accuracy, which made up the defects of the time-consuming and the low-accuracy of the random search algorithm in the grid search when selecting parameters. At last, this paper applied the random forest algorithm combing with simulated annealing to the housing prices evaluation, and formed a new evaluation algorithm. Compared with the traditional random forest price estimation algorithms, the results show that the proposed algorithm reduces the error value, increases the goodness of fit value, and improves markedly the accuracy of the evaluation.