TWI759702B - Real estate appraisal system - Google Patents

Abstract

The present disclosure provides a real estate appraisal system including a system host and an application host coupled to the system host. The system host stores an appraisal model, wherein the appraisal model includes an ensemble learning model. The ensemble learning model includes a first model and a second model and the ensemble learning model receives an input data and generates an output data. The system host receives multiple refresh data in an time interval and input the refresh data into the appraisal model to perform a validity verification. When an output result of the refresh data does not pass the validity verification, the system host does not adjust multiple parameters of the appraisal model and remodel the appraisal model according to the refresh data. When the remodeled appraisal model passes the validity verification, the system host deploys the appraisal model to the application host to perform an appraisal operation.

Description

real estate appraisal system

The present disclosure relates to a real estate appraisal system, and in particular, to a real estate appraisal system that automatically updates the appraisal model.

For the general public, it is expensive to buy a house. Therefore, several target items are often selected for condition, price confirmation and comparison before the transaction. For example, the public can confirm the reasonableness of the price through various channels such as real-price login websites, real estate brokers, and property owners' quotations. However, this kind of work is time-consuming and labor-intensive, and the inquiries may also have hidden evils or high prices. In order to solve the above-mentioned problem of buying a house, an online real estate appraisal system has been launched. However, most of the existing online real estate appraisal systems have a large gap between the appraisal result and the actual price due to insufficient reference data. Therefore, how to provide a high-precision real estate appraisal system is the goal of those skilled in the art.

In view of this, the present disclosure provides a real estate appraisal system that can automatically update the appraisal model to provide more accurate appraisal results.

The present disclosure provides a real estate appraisal system, including a system host and an application host coupled to the system host. The system host stores the appraisal model, wherein the appraisal model includes an integrated learning model, the integrated learning model includes a first model and a second model, and the integrated learning model receives input data and generates output data. The system host receives multiple updates at time intervals and inputs the updates into the appraisal model for validity verification. When the output result of the updated data fails the validity verification, the system host does not adjust multiple parameters of the appraisal model and re-models the appraisal model according to the updated data. When the remodeled appraisal model passes the validity verification, the system host deploys the appraisal model to the application host to provide appraisal operations.

Based on the above, the real estate appraisal system of the present disclosure utilizes the ensemble learning model as the appraisal model. The system host receives updates, such as monthly, and enters the updates into the appraisal model for validity verification. If the validity verification fails, the system host does not temporarily adjust multiple parameters of the appraisal model, but first re-models the appraisal model according to the updated data. If the remodeled appraisal model passes the validity verification, the system host can deploy the appraisal model to the application host to provide appraisal operations.

FIG. 1 is a block diagram of a real estate appraisal system according to an embodiment of the present disclosure.

Referring to FIG. 1 , a real estate appraisal system 100 according to an embodiment of the present disclosure includes a system host 110 and an application host 120 . The application host 120 is coupled to the system host 110 through a wired or wireless network.

In one embodiment, the system host 110 stores an appraisal model, wherein the appraisal model includes an ensemble learning model. The integrated learning model includes a first model and a second model. The first model is, for example, a random forest model and the second model is, for example, an eXtreme Gradient Boosting (XgBoost) model. An ensemble learning model receives input data and produces output data. Multiple model outputs in the ensemble learning model can form a stacking model of Linear Regression. The input data may include credit information, actual price registration data, point of interest (POI) data, and economic indicator data and other real estate-related data, and the output data are, for example, real estate appraisal results. The system host 110 receives multiple updates at time intervals (eg, every month) and enters the updates into the appraisal model for validity verification. When the output result of the updated data fails the validity verification, the system host 110 does not adjust a plurality of parameters of the appraisal model and re-models the appraisal model according to the updated data. When the remodeled appraisal model passes the validity verification, the system host 110 deploys the appraisal model to the application host 120 to provide appraisal operations, and updates the appraisal model Application Program Interface (API) The model file for the front desk to call the latest appraisal results. When the output result of the updated data does not pass the validity verification of the remodeled appraisal model, the system host 110 updates a plurality of parameters corresponding to the appraisal model and applies these parameters to the appraisal model.

In one embodiment, the system host 110 may update a plurality of parameters corresponding to the appraisal model according to a grid search algorithm or a genetic algorithm. Taking the grid search algorithm as an example, the grid search algorithm can try every possibility in a round-robin manner among all the candidate parameter combinations, and take the best performance parameter combination as the final result. In one embodiment, both the random forest model and the extreme gradient boosting model can use a grid search algorithm to automatically find the best parameter combination of the appraisal model. Table 1 shows the model, parameters and parameter descriptions of grid search.

Table I Model parameter Parameter Description Random Forest Model n_estimators Number of trees in random forest max_depth Maximum depth of tree in random forest min_samples_split Minimum number of samples required for branching min_samples_leaf Minimum sample size for leaf nodes Extreme Gradient Boosting Model eta learning rate max_depth maximum depth of tree subsample Sample sampling ratio colsample_bytree Variable sampling ratio gamma Lower bound of loss function value min_child_weight Child node sample weights and

On the other hand, the system host 110 can also update a plurality of parameters corresponding to the appraisal model according to the genetic algorithm. Using genetic algorithms to determine the optimal parameters is much more efficient than grid search.

In one embodiment, the update data includes a plurality of real estate data, and when the difference between the output results generated by the real estate data input into the appraisal model and the plurality of predetermined appraisal results corresponding to the real estate data is less than or equal to a threshold value, the output results pass the validation. degree verification. When the difference between the output result and the predetermined appraisal result is greater than the threshold value, the output result does not pass the validity verification.

For example, the system host 110 may calculate the Mean Absolute Percentage Error (MAPE) and the hit rate of the corresponding output result. When the mean absolute error percentage is less than or equal to a predetermined percentage (eg, 15%) and the error is within a first percentage (eg, 10%), the hit rate is greater than or equal to a second predetermined percentage (eg, 50%) and the error is within a second If the hit rate within the percentage (for example, 20%) is greater than or equal to the third predetermined percentage (for example, 80%), the output result passes the validity verification. Through the double verification of mean absolute error percentage and hit rate, the accuracy of validity verification can be greatly increased.

，其中y為鑑價結果實際值且ŷ為鑑價結果預測值且n為樣本數。命中率是指在特定誤差範圍內，預測值落於該區間的機率，命中率越高，表示預測值接近實際值的機率越高。命中率的計算公式為

，

，其中y為鑑價結果實際值且ŷ為鑑價結果預測值且

為信心水準且N為測試樣本數且n為命中區間樣本數。The mean absolute error percentage is used to measure the ratio of the difference between the actual value and the predicted value to the actual value. The formula for calculating the mean absolute error percentage is

, where y is the actual value of the appraisal result and ŷ is the predicted value of the appraisal result and n is the number of samples. The hit rate refers to the probability that the predicted value falls within this range within a specific error range. The higher the hit rate, the higher the probability that the predicted value is close to the actual value. The formula for calculating the hit rate is

,

, where y is the actual value of the appraisal result and ŷ is the predicted value of the appraisal result and

is the confidence level and N is the number of test samples and n is the number of hit interval samples.

To sum up, the real estate appraisal system of the present disclosure utilizes the ensemble learning model as the appraisal model. The system host receives updates, such as monthly, and enters the updates into the appraisal model for validity verification. If the validity verification fails, the system host does not temporarily adjust multiple parameters of the appraisal model, but first re-models the appraisal model according to the updated data. If the remodeled appraisal model passes the validity verification, the system host can deploy the appraisal model to the application host to provide appraisal operations. If the re-modeled appraisal model still fails to pass the validity verification, the parameters of the appraisal model are updated according to the grid search algorithm or the genetic algorithm to improve the chance of passing the validity verification.

Although the present disclosure has been disclosed above with examples, it is not intended to limit the present disclosure. Anyone with ordinary knowledge in the technical field may make some changes and modifications without departing from the spirit and scope of the present disclosure. The scope of protection of the present disclosure shall be determined by the scope of the appended patent application.

100: Real estate appraisal system 110: System host 120: Application Host

FIG. 1 is a block diagram of a real estate appraisal system according to an embodiment of the present disclosure.

100: Real estate appraisal system

110: System host

120: Application Host

Claims (9)

A real estate appraisal system, comprising: a system host; and an application host coupled to the system host, wherein the system host stores an appraisal model, wherein the appraisal model includes an integrated learning model, the integrated learning model A first model and a second model are included and the integrated learning model receives an input data and generates an output data; the system host receives a plurality of update data at a time interval and inputs the updated data into the appraisal model for performing a validity verification; when an output result of the updated data fails the validity verification, the system host does not adjust a plurality of parameters of the appraisal model and re-models the appraisal model according to the updated data; when When the remodeled appraisal model passes the validity verification, the system host deploys the appraisal model to the application host to provide a appraisal operation; and when the output of the updated data does not pass the reappraisal When the validity of the modeled appraisal model is verified, the system host updates the parameters corresponding to the appraisal model and applies the parameters to the appraisal model. The real estate appraisal system according to claim 1, wherein the update data includes a plurality of real estate data, and when the real estate data is input into the appraisal model, the output result generated by the appraisal model corresponds to a plurality of predetermined appraisal results corresponding to the real estate data The difference is less than or equal to a threshold value, then the output result passes the validity verification, when the output result is different from The difference between the predetermined evaluation results is greater than the threshold value, and the output result does not pass the validity verification. The real estate appraisal system according to claim 2, wherein the system host calculates a mean absolute error percentage and a hit rate corresponding to the output result, when the mean absolute error percentage is less than or equal to a predetermined percentage and the error is within the first percentage If the hit rate within the ratio is greater than or equal to the second predetermined percentage and the hit rate within the second percentage is greater than or equal to the third predetermined percentage, the output result passes the validity verification. The real estate appraisal system of claim 3, wherein the mean absolute percentage error is

, where y is the actual value of the appraisal result and

is the predicted value of the appraisal result and n is the number of samples, the hit rate is

, y _i - y _{i (α)}

y _i + y _{i (α)} , where y is the actual value of the appraisal result and

is the predicted value of the appraisal result and α is the confidence level and N is the number of test samples and n is the number of hit interval samples. The real estate appraisal system of claim 1, wherein the system host updates the parameters corresponding to the appraisal model according to a grid search algorithm or a genetic algorithm. The real estate appraisal system of claim 1, wherein the first model is a random forest model. The real estate appraisal system of claim 1, wherein the second model is an extreme gradient boosting (eXtreme Gradient Boosting, XgBoost) model. The real estate appraisal system according to claim 1, wherein the input data includes a credit information, a real-price registration data, a point of interest data and an economic indicator data. The real estate appraisal system of claim 1, wherein the output data includes a real estate appraisal result.

Images