Method and device for generating house resource task

Abstract

The invention provides a method and device for generating a house resource task, and the method comprises: inputting a current feature set of a target house resource into a preset house resource prediction model, and obtaining a first estimation value; based on the current feature set and the first estimated value, obtaining an importance value of each feature through a model interpretation tool;determining a first disturbance feature based on the importance value of each feature, and performing first disturbance on the first disturbance feature to form a first adjustment feature set; inputting the first adjustment feature set into a house resource prediction model to obtain a second estimated value; comparing the first pre-estimated value with the second pre-estimated value, and if the second pre-estimated value is smaller than or equal to the first pre-estimated value, filtering the target house resource; and if the second estimated value is greater than the first estimated value, generating a house resource task based on the first disturbance. According to the method and the device, the interpretation of the prediction model can be provided, and the working efficiency and accuracy in the field of real estate are improved.

Description

technical field [0001] The invention relates to the field of intelligent house source processing, in particular to a method and device for generating house source tasks. Background technique [0002] In today's real estate industry, it is necessary to know which houses are easy to sell, which are highly competitive, which have defects, and what factors affect the probability of house transaction, and how these factors affect the prediction results of house transaction probability, etc. Information is crucial to real estate agents or users. [0003] However, what the current industry can do is to give a prediction result through a computer model, such as a model that predicts the probability of a house being sold in the next N days. The input features of the model are various characteristics related to the house, and the output is the The prediction result of the probability that the listing will be sold within N days. However, brokers or users cannot know the explanation o...

AI Technical Summary

Problems solved by technology

However, brokers or users cannot know the explanation of the prediction results, that is, how each feature of the house source affects the prediction results, which will lead to a decrease in the trust of brokers or users in the prediction results, and thus cannot Make targeted recommendations for brokers and formulate plans to increase the probability of closing a house

Images