Artificial intelligence model surveillance system
Systems and methods for property valuation using automated valuation models are described herein, which involve a complex series of workflows for data processing, automated valuation modeling, and error detection—each of which are capable of leveraging machine learning algorithms/techniques and can be combined and operated together in a specific manner in order to improve data availability and data quality, improve the accuracy of generated property value estimates, improve the performance of the models over time, and allow the property value estimation to be adapted to various use cases. Estimation of property value may involve an ensemble model which reconciles the outputs of different sub-models based on a use case, with the sub-models using different approaches that have their own strengths and weaknesses.
1 . A computer-implemented method comprising:
obtaining model data input to a first machine learning model in a plurality of machine learning models used in a property valuation workflow or output from the first machine learning model;
applying the model data as an input to at least one of a plurality of machine learning anomaly detection models, wherein the at least one of the plurality of machine learning anomaly detection models is different than the first machine learning model;
detecting an anomaly in the model data applied as the input to the at least one of the plurality of machine learning anomaly detection models based on an output of the at least one of the plurality of machine learning anomaly detection models;
generating a suggestion to modify at least some of the model data based on a profile associated with the detected anomaly, wherein the suggestion comprises an identification of the detected anomaly;
presenting the suggestion to a user for review;
modifying the at least some of the model data in accordance with the generated suggestion; and
re-training the first machine learning model using the modified at least some of the model data.
2 . The method of claim 1 , wherein modifying the at least some of the model data associated with the execution of the first machine learning model further comprises modifying the at least some of the model data automatically without user intervention.
3 . The method of claim 1 , wherein re-training the first machine learning model improves performance of the first machine learning model.
4 . The method of claim 3 , wherein re-training the first machine learning model further comprises re-training the first machine learning model automatically without user intervention.
5 . The method of claim 1 , wherein the model data associated with execution of the first machine learning model comprises at least one of model input data, model exhaust data, exhaust generated attributes, or complimentary data.
6 . The method of claim 1 , wherein the profile associated with the detected anomaly includes a cause for the detected anomaly.
7 . The method of claim 6 , wherein the profile associated with the detected anomaly includes an approach for correcting or preventing the detected anomaly.
8 . The method of claim 1 , further comprising monitoring individual performance of the first machine learning model on a continual basis.
9 . A system comprising:
one or more computers; and
computer storage media storing instructions that when executed by the one or more computers, cause the one or more computers to:
obtain model data input to a first machine learning model in a plurality of machine learning models used in a property valuation workflow or output from the first machine learning model;
apply the model data as an input to at least one of a plurality of machine learning anomaly detection models, wherein the at least one of the plurality of machine learning anomaly detection models is different than the first machine learning model;
detect an anomaly in the model data applied as the input to the at least one of the plurality of machine learning anomaly detection models based on an output of the at least one of the plurality of machine learning anomaly detection models;
generate a suggestion to modify at least some of the model data based on a profile associated with the detected anomaly, wherein the suggestion comprises an identification of the detected anomaly;
present the suggestion to a user for review;
modify the at least some of the model data in accordance with the generated suggestion; and
re-train the first machine learning model using the modified at least some of the model data.
10 . The system of claim 9 , wherein the instructions, when executed, further cause the one or more computers to modify the at least some of the model data automatically without user intervention.
11 . The system of claim 9 , wherein the instructions, when executed, further cause the one or more computers to re-train the first machine learning model automatically without user intervention.
12 . The system of claim 9 , wherein the model data associated with execution of the first machine learning model comprises at least one of model input data, model exhaust data, exhaust generated attributes, or complimentary data.
13 . The system of claim 9 , wherein the profile associated with the detected anomaly includes a cause for the detected anomaly.
14 . The system of claim 13 , wherein the profile associated with the detected anomaly includes an approach for correcting or preventing the detected anomaly.
15 . Non-transitory computer storage media storing instructions that when executed by a system of one or more computers, cause the one or more computers to perform operations comprising:
obtaining model data input to a first machine learning model in a plurality of machine learning models used in a property valuation workflow or output from the first machine learning model;
applying the model data as an input to at least one of a plurality of machine learning anomaly detection models, wherein the at least one of the plurality of machine learning anomaly detection models is different from the first machine learning model;
detecting an anomaly in the model data applied as the input to the at least one of the plurality of machine learning anomaly detection models based on an output of the at least one of the plurality of machine learning anomaly detection models;
generating a suggestion to modify at least some of the model data based on a profile associated with the detected anomaly, wherein the suggestion comprises an identification of the detected anomaly;
presenting the suggestion to a user for review;
modifying the at least some of the model data in accordance with the generated suggestion; and
re-training the first machine learning model using the modified at least some of the model data.
16 . The non-transitory computer storage media of claim 15 , wherein the first machine learning model, a second machine learning model, and a third machine learning model comprise, at least in part, a plurality of sub-models and an ensemble model, and wherein the ensemble model is applied to outputs of the plurality of sub-models.