WO2023053112A1

WO2023053112A1 - System and method for prediction of residential real-estate values

Info

Publication number: WO2023053112A1
Application number: PCT/IL2022/051017
Authority: WO
Inventors: Peleg DAVIDOVITZ
Original assignee: Propdo Ltd.
Priority date: 2021-09-30
Filing date: 2022-09-22
Publication date: 2023-04-06

Abstract

A system for computer-implemented predictive modeling for residential real-estate property comprising: A computer system containing a processor, configured to aggregate sample data regarding multiple factors associated with a particular residential property; perform iterative analysis on a sample dataset using machine learning to construct a predictive model; validate said predictive model using a validation dataset; creating a model to evaluate the current value of residential properties and to predict the future value of said properties.

Description

SYSTEM AND METHOD FOR PREDICTION OF RESIDENTIAL REAL-ESTATE

VALUES

Field of invention

The present disclosure relates to generally to real-estate valuation tools, and more specifically to a system and method for valuation and prediction of residential real-estate.

Background of invention

The field of property valuation is still mostly done by a real-estate appraiser. The problem being that often two different appraisers will give two different results. As such, there is a need for an automatic system that can predict, using multiple data-points, the current value, as well as future values of residential properties.

US 16/505259 - “predictive machine learning models” specifies application specifically deals with ways to train machine learning modules. The patent refers to training machine learning algorithms with real estate information and arriving at predictions based on those models. The patent refers to some methods to reduce the inaccuracies of said predictions, however it does not specifically mention the problem of predicting future value of real estate, just helps with risk management, mortgage, and insurance rates, etc. This patent mostly deals with predictions the likelihood of a mortgage is attached to a specific property, as well as the method of determining machine learning model performance, etc. This patent does not deal with specific predictions for residential real-estate valuations.

US 16/383250 - “Method and system for predicting and indexing real estate demand and pricing.” This patent has several embodiments in this application which closely relate to the current invention. Embodiments of this invention include predictive models for property based on geographic location, salary, taxation, employment, etc. in geographical areas. However, this patent’s prediction model is aimed at predicting demand for property, and not for the valuation - or for assessing future investment or mortgage refinancing, etc. US 16/391945 - “System and method for generating value prediction of commercial realestate” This patent is set specifically to provide predictions to commercial real-estate. The patent specifically indicates predictions regarding future value predictions for commercial real-estate. It should be noted that machine learning algorithms used to determine predictions for commercial real-estate will use different variables then the ones used to evaluate residential real-estate and will naturally produce different results.

KR 10201901235122 - “System for providing real estate transaction information and price prediction information, and a method thereof’ - The patent describes an algorithm trained with public information regarding real-estate prices and data and provides forecasting information regarding the real-estate that may include future costs. However, this patent seems to only use public information to predict the future values of properties. It does not allow for data that not be available online (apartment innovations, possible future area plans, apartment view, etc.). This added data can completely change the value prediction of a property.

In view of the above, there is still an unmet long-felt need for a system that will provide accurate value predictions for residential real-estate. This need is both for those who are considering buying real-estate as a primary residence and need to plan out their mortgage, and for investors who want to plan the finances for future rent and mortgage payments.

Summary of the invention:

It is thus one object of the present invention to disclose a method useful for a system featuring a computer-implemented predictive modeling for residential real-estate property comprising a computer system containing at least one processor, wherein said processor aggregate sample data regarding multiple factors associated with a particular residential property; perform iterative analysis on a sample dataset using machine learning to construct a predictive model; validate said predictive model using a validation dataset; creating a model to evaluate the current value of residential properties and to predict the future value of said properties. The system machine learning module will be trained using categories of a sample dataset used in the machine learning training phase include at least one datapoints of population, employment density, average rent price, average sale price, environmental scores, socioeconomic factors, financial, employment, education, statistical data for district, city and neighborhood, amenities, city services, activities, transportation, and pollution scores.

The system can further increase its accuracy by including subjective datapoints, including view, internal decorations, internal amenities, internal layout, disturbances which the user can manually enter. The system then includes said subjective datapoints in the prediction valuation of a real-estate property. The system can use said data to then further calculate mortgage values and returns, current and future rent values.

It is thus one object of the present invention to disclose a method useful for a method for generating prediction of residential real-estate property where a computer system will receive an address associated with a residential property, extract data associated with the residential property, and then analyzing said data, wherein the analysis includes matching residential real-estate factors, and generating at least one value prediction of the residential property based on the analysis.

The method machine learning module will be trained using categories of the data used to calculate and predict values of a residential property include population, employment density, average rent price, average sale price, environmental scores, socioeconomic factors, financial, employment, education, statistical data for district, city and neighborhood, amenities, city services, activities, transportation, and pollution scores.

The method can further increase the accuracy of the output by including at least one of these subjective datapoints: view, internal decorations, internal amenities, internal layout, disturbances. The method will then output a valuation that includes the subjective datapoints. Brief description of the figures

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serves to explain the principles of the invention.

Fig. 1 depicting a schematic representation of the cloud based system;

Fig. 2 depicting a schematic presentation the transaction database;

Fig. 3 depicting a flow chart of the machine learning training method;

Fig. 4 depicting a flow chart recalculating prediction based on user input;

Fig. 5 depicting a flow chart of user interaction with the system;

Fig. 6 depicting a sample graph output of current and predicted property values.

Detailed description of preferred embodiments

In this document, the term “ETL” stands for Extract-Transform-Load, a layer for concentrating data from various sources into one organized database, thereafter, used by the machine learning modules for training, valuation, and prediction.

Fig. 1 describes a basic schematic of the system, wherein a user connects through a device (101) to a cloud service (102) that holds both a server (104) and the transactions database (103). The user enters his query into the system, which is then received by the server, containing at least one processor and memory module. The server runs the machine learning algorithms, using data taken from the transactions database to produce the value predictions, and returns the data back to the user device.

Fig. 2 describes the data the system collects from multiple sources, including city services and data, such as education, amenities, etc. (201), statistical data such as district and city statistical data (202), both combined as hyper-location data (205). Additionally, indices are collected (206) including environmental, social, socioeconomic data (203) as well as census data (207) that includes employment density data, average sale and rent prices, etc. (204). All this data goes through and ETL layer to standardize the data (208) and store it in the database (210). Extra property transactions are also added into the database (209), these are updated for future calculations of cost assessments.

Fig. 3 describes the machine learning training process. The transactions database (301) holds both a training set and a validation set of data (303, 302). The system sends out data from the training set to the machine learning module (304) and uses hyperparameter optimization (305). These conclude in an Al output model (307). The Al output model is validated with the validation set (302). Finally, an Al model for quality score (308) is tested against an external test set (306).

Fig. 4 describes the method of receiving extra variables from a user regarding objective and non-objective information about a particular property. In this case the system starts with a basis price prediction (401) which is based on parameters noted above in Fig. 2. To this prediction the user adds subjective input (such as view, disbursements, etc. (404)), TAMA, or building renovations (402), property condition (403), potential building rights (406) and area statistics (407). All of these are recalculated using the base prediction price (408) and a new output based on said data is given to the user (409).

Fig. 5 describes the user interaction with the system, as well as a subset of the potential system output. The user inputs the search he desires (501). The system calculates a price prediction (506), checks for future area plans (505), and checks future rent values (504). The price prediction generates a current value estimation (510), renovation impact (509) and a 25-year forecast (511). The system then generates from all the data an expected investment return (513) and a mortgage calculation or refinancing costs (512).

Fig. 6 shows an example graph of the result a user will get from the system. The graph showing the valuation of increase (or decrease) of the residential property value over a period of 25 years, including optimal, realistic, and pessimistic predictions over that time period.

It is important to note that the embodiments disclosed herein are only examples of the many advantageous uses of the innovative teachings herein. In general, statements made in the specification of the present application do not necessarily limit any of the various claimed embodiments. Moreover, some statements may apply to some inventive features but not to others. In general, unless otherwise indicated, singular elements may be in plural and vice versa with no loss of generality. In the drawings, like numerals refer to like parts through several views.

The following description is provided, alongside all chapters of the present invention, so as to enable any person skilled in the art to make use of the invention and sets forth the best modes contemplated by the inventor of carrying out this invention. Various modifications, however, are adapted to remain apparent to those skilled in the art, since the generic principles of the present invention have been defined specifically to provide a solution to the problem of predicting residential real estate costs, over a variety of timelines, based on both objective and subjective datapoints.

The system disclosed is configured to receive a full address of a residential real-estate (apartment, house) and output a general valuation for said property. The user can then modify the valuation based on specific data about said property (e.g. number of rooms, renovations done, building plans in the neighborhood, etc.). The user will then get a valuation, with a

The present invention provides a method to teach and test a machine learning module and optimize the results it generates. The teaching phase consists of giving the module a test data set to process, get values, and test these values against a validation data set. The system further checks the machine learning algorithms against external test sets to make sure the system produces the best results.

Claims

1. A system for computer- implemented predictive modeling for residential realestate property comprising:

A computer system containing at least one processor, wherein said processor is configured to aggregate sample data regarding multiple factors associated with a particular residential property; perform iterative analysis on a sample dataset using machine learning to construct a predictive model; validate said predictive model using a validation dataset; creating a model to evaluate the current value of residential properties and to predict the future value of said properties.

2. The system described in claim 1, wherein categories of the sample dataset used in the machine learning training phase include at least one of these datapoints: Population, employment density, average rent price, average sale price, environmental scores, socioeconomic factors, financial, employment, education, statistical data for district, city and neighborhood, amenities, city services, activities, transportation, and pollution scores.

3. The system described in claim 1, wherein the user can include one of these subjective datapoints: view, internal decorations, internal amenities, internal layout, disturbances.

4. The system described in claim 4, wherein the system includes said subjective datapoints in the output prediction value.

5. The system described in claim 1, wherein the system calculates mortgage costs and returns

6. The system described in claim 1, wherein the system calculates current rent values

7. The system described in claim 1, wherein the system calculates future rent values

8. A method for generating prediction of residential real-estate property comprising:

7 Receiving an address associated with a residential property; extracting data associated with the residential property; analyzing said data, wherein the analysis includes matching residential real-estate factors, and generating at least one value prediction of the residential property based on the analysis. The method described in claim 8, wherein categories of the data used to calculate and predict values of a residential property include at least one of:

Population, employment density, average rent price, average sale price, environmental scores, socioeconomic factors, financial, employment, education, statistical data for district, city and neighborhood, amenities, city services, activities, transportation, and pollution scores. The method described in claim 8, wherein the user can include one of these subjective datapoints: view, internal decorations, internal amenities, internal layout, disturbances. The method described in claim 8, wherein the system includes said subjective datapoints in the output prediction value.

8