Machine learning models supported by big data have been practiced and applied in many ways in recent years, and as a representative technology of artificial intelligence, machine learning models have been proved to be able to perform well in many predictive problems such as economics and management. This paper explores the practice in the problem of residential value assessment by using the more popular machine learning models. The Chain Home platform offers publicly available, granular data on residential property transactions, including variables such as location, area, layout, and pricing. The dataset from November 22, 2024, was selected to provide a consistent time snapshot of the housing market, facilitating reliable model training and evaluation. After that, it further compares the performance of linear regression, random forest algorithm, extreme gradient boosting tree, lightweight gradient boosting tree, classification boosting tree and other algorithms on asset pricing. The empirical results show that the machine learning algorithms can be relatively effective in assessing and pricing residential properties according to their characteristics, and the error between the predicted price and the actual price of the asset appraisal model based on the extreme boosted tree algorithm is much smaller, with an average error of about 17%. This paper attempts to introduce machine learning into the field of asset evaluation, which helps to promote the cross-fertilization research of artificial intelligence and traditional economics problems, and provides reference for promoting the application of artificial intelligence.