CN112927013B - Asset value prediction model construction method and asset value prediction method - Google Patents

Asset value prediction model construction method and asset value prediction method Download PDF

Info

Publication number
CN112927013B
CN112927013B CN202110206060.9A CN202110206060A CN112927013B CN 112927013 B CN112927013 B CN 112927013B CN 202110206060 A CN202110206060 A CN 202110206060A CN 112927013 B CN112927013 B CN 112927013B
Authority
CN
China
Prior art keywords
model
asset value
predicted
training data
value corresponding
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202110206060.9A
Other languages
Chinese (zh)
Other versions
CN112927013A (en
Inventor
陈绍真
张兴华
何清素
张程
王建文
何金霖
张炀
周峰
俞果
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guowang Xiongan Finance Technology Group Co ltd
State Grid Digital Technology Holdings Co ltd
Original Assignee
Guowang Xiongan Finance Technology Group Co ltd
State Grid Digital Technology Holdings Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guowang Xiongan Finance Technology Group Co ltd, State Grid Digital Technology Holdings Co ltd filed Critical Guowang Xiongan Finance Technology Group Co ltd
Priority to CN202110206060.9A priority Critical patent/CN112927013B/en
Publication of CN112927013A publication Critical patent/CN112927013A/en
Application granted granted Critical
Publication of CN112927013B publication Critical patent/CN112927013B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0283—Price estimation or determination
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0201—Market modelling; Market analysis; Collecting market data

Abstract

The application discloses an asset value prediction model construction method and an asset value prediction method, which are characterized in that training data, corresponding actual asset values, verification data, corresponding actual asset values and a preset neural network model are utilized to construct an asset value prediction model, so that the constructed asset value prediction model has good asset value prediction performance; then, after the value influence characteristics of the target object (for example, a power distribution network) are acquired, the value influence characteristics of the target object are directly input into a constructed asset value prediction model, so that the asset value prediction model can accurately predict the predicted asset value of the target object according to the value influence characteristics of the target object, and the prediction accuracy of the asset value is improved.

Description

Asset value prediction model construction method and asset value prediction method
Technical Field
The application relates to the technical field of data processing, in particular to an asset value prediction model construction method and an asset value prediction method.
Background
In securitizing a target object (e.g., a power distribution network), asset value predictions may be made for the base assets of the target object. However, due to the low accuracy of the existing asset value prediction process, how to accurately predict the asset value of the base asset is a technical problem to be solved.
Disclosure of Invention
In order to solve the technical problems in the prior art, the application provides an asset value prediction model construction method and an asset value prediction method, which can accurately predict the asset value of a base asset, so that the prediction accuracy of the asset value can be effectively improved.
In order to achieve the above object, the technical solution provided by the embodiments of the present application is as follows:
the embodiment of the application provides a method for constructing an asset value prediction model, which comprises the following steps:
acquiring training data, actual asset values corresponding to the training data, verification data and actual asset values corresponding to the verification data;
inputting the training data and the verification data into a preset neural network model to obtain a first predicted asset value corresponding to the training data and a predicted asset value corresponding to the verification data output by the preset neural network model;
determining model performance evaluation characteristics corresponding to the current model updating times according to the first predicted asset value corresponding to the training data, the actual asset value corresponding to the training data, the predicted asset value corresponding to the verification data and the actual asset value corresponding to the verification data;
And updating the preset neural network model by utilizing the first predicted asset value corresponding to the training data and the actual asset value corresponding to the training data, continuously executing the input of the training data and the verification data into the preset neural network model, and obtaining the first predicted asset value corresponding to the training data and the predicted asset value corresponding to the verification data output by the preset neural network model and subsequent steps until the predicted asset value reaches a preset stopping condition according to the model performance evaluation characteristics, and constructing an asset value prediction model.
In one possible implementation manner, the constructing an asset value prediction model according to the model performance evaluation feature includes:
determining the model updating times corresponding to the model overfitting initial characterization points according to the model performance evaluation characteristics; the model overfitting initial characterization points are used for characterizing the overfitting phenomenon in the training process of the preset neural network model;
determining the updating times of the target model according to the model updating times corresponding to the model overfitting initial characterization points;
and constructing an asset value prediction model according to the updating times of the target model.
In one possible implementation manner, if the model performance evaluation feature includes a predicted performance characterization value corresponding to the training data and a predicted performance characterization value corresponding to the verification data, the determining process of the model over-fitting the initial characterization point includes:
generating a predicted performance change curve corresponding to the training data according to the predicted performance characterization value corresponding to the training data; the predicted performance change curve corresponding to the training data is used for describing the change trend of the predicted performance representation value corresponding to the training data along with the model updating times of the preset neural network model;
generating a predicted performance change curve corresponding to the verification data according to the predicted performance characterization value corresponding to the verification data; the predicted performance change curve corresponding to the verification data is used for describing the change trend of the predicted performance characterization value corresponding to the verification data along with the model updating times of the preset neural network model;
and determining a model overfitting initial characterization point meeting a preset overfitting initial condition according to the predicted performance change curve corresponding to the training data and the predicted performance change curve corresponding to the verification data.
In one possible implementation manner, the determining the number of updating the target model according to the number of updating the model corresponding to the model overfitting initial characterization point includes:
subtracting 1 from the model updating times corresponding to the model fitting initial characterization points to obtain target model updating times;
or,
and determining the model updating times corresponding to the model fitting initial characterization points as target model updating times.
In one possible implementation manner, the constructing an asset value prediction model according to the update times of the target model includes:
determining an untrained preset neural network model as a model to be used;
inputting the training data into the model to be used to obtain a second predicted asset value corresponding to the training data output by the model to be used;
updating the model to be used according to the second predicted asset value corresponding to the training data and the actual asset value corresponding to the training data, and continuously executing the step of inputting the training data into the model to be used to obtain the second predicted asset value corresponding to the training data output by the model to be used and subsequent steps until the model to be used is determined to be an asset value prediction model when the current updating times of the model to be used is determined to reach the target model updating times.
In one possible implementation manner, if the model performance evaluation feature includes a predicted performance characterization value corresponding to the training data and a predicted performance characterization value corresponding to the verification data, the determining process of the predicted performance characterization value corresponding to the training data includes:
when the number of the training data is M 1 When M is to 1 First predicted asset value corresponding to the training data and the M 1 Determining an average absolute error between actual asset values corresponding to the training data as a predicted performance characterization value corresponding to the training data;
and/or the number of the groups of groups,
the determining process of the predicted performance characterization value corresponding to the verification data comprises the following steps:
when the number of the verification data is M 2 When M is to 2 Predicted asset value corresponding to individual validation data and the M 2 And determining the average absolute error between the actual asset values corresponding to the verification data as a predicted performance characterization value corresponding to the verification data.
In one possible implementation manner, the preset neural network model includes a first hidden layer and a second hidden layer, input data of the second hidden layer includes output data of the first hidden layer, the first hidden layer includes 16 nodes, the second hidden layer includes 8 nodes, and an activation function of each node is a hyperbolic tangent activation function.
The embodiment of the application also provides an asset value prediction method, which comprises the following steps:
acquiring value influence characteristics of a target object;
inputting the value influence characteristics of the target object into a pre-constructed asset value prediction model to obtain the predicted asset value of the target object output by the asset value prediction model; the asset value prediction model is built by any implementation mode of the asset value prediction model building method provided by the embodiment of the application.
The embodiment of the application also provides an asset value prediction model construction device, which comprises the following steps:
the first acquisition unit is used for acquiring training data, actual asset values corresponding to the training data, verification data and actual asset values corresponding to the verification data;
the first prediction unit is used for inputting the training data and the verification data into a preset neural network model to obtain a first predicted asset value corresponding to the training data and a predicted asset value corresponding to the verification data output by the preset neural network model;
the performance evaluation unit is used for determining model performance evaluation characteristics corresponding to the current model updating times according to the first predicted asset value corresponding to the training data, the actual asset value corresponding to the training data, the predicted asset value corresponding to the verification data and the actual asset value corresponding to the verification data;
And the model updating unit is used for updating the preset neural network model by utilizing the first predicted asset value corresponding to the training data and the actual asset value corresponding to the training data, returning to the first predicting unit to continuously execute the input of the training data and the verification data into the preset neural network model to obtain the first predicted asset value corresponding to the training data and the predicted asset value corresponding to the verification data output by the preset neural network model, and constructing an asset value predicting model according to the model performance evaluation characteristics after the preset stopping condition is determined to be reached.
The embodiment of the application also provides an asset value prediction device, which comprises:
a second acquisition unit configured to acquire a value influence feature of the target object;
the second prediction unit is used for inputting the value influence characteristics of the target object into a pre-constructed asset value prediction model to obtain the predicted asset value of the target object output by the asset value prediction model; the asset value prediction model is built by any implementation mode of the asset value prediction model building method provided by the embodiment of the application.
The embodiment of the application also provides equipment, which comprises a processor and a memory:
the memory is used for storing a computer program;
the processor is configured to execute any implementation mode of the asset value prediction model construction method provided by the embodiment of the present application according to the computer program, or execute any implementation mode of the asset value prediction method provided by the embodiment of the present application.
The embodiment of the application also provides a computer readable storage medium, which is used for storing a computer program, wherein the computer program is used for executing any implementation mode of the asset value prediction model construction method provided by the embodiment of the application or executing any implementation mode of the asset value prediction method provided by the embodiment of the application.
The embodiment of the application also provides a computer program product, which when being run on a terminal device, causes the terminal device to execute any implementation mode of the asset value prediction model construction method provided by the embodiment of the application or execute any implementation mode of the asset value prediction method provided by the embodiment of the application.
Compared with the prior art, the embodiment of the application has at least the following advantages:
in the technical scheme provided by the embodiment of the application, the training data and the corresponding actual asset value thereof, the verification data and the corresponding actual asset value thereof and the preset neural network model are utilized to construct an asset value prediction model, so that the constructed asset value prediction model has better asset value prediction performance; then, after the value influence characteristics of the target object (for example, a power distribution network) are acquired, the value influence characteristics of the target object are directly input into a constructed asset value prediction model, so that the asset value prediction model can accurately predict the predicted asset value of the target object according to the value influence characteristics of the target object, and the prediction accuracy of the asset value is improved.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments described in the present application, and other drawings may be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flowchart of a method for constructing an asset value prediction model according to an embodiment of the present application;
fig. 2 is a schematic structural diagram of a node according to an embodiment of the present application;
FIG. 3 is a flowchart of a method for asset value prediction according to an embodiment of the present application;
FIG. 4 is a schematic structural diagram of an asset value prediction model construction device according to an embodiment of the present application;
fig. 5 is a schematic structural diagram of an asset value predicting device according to an embodiment of the present application.
Detailed Description
In order to make the present application better understood by those skilled in the art, the following description will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
In order to facilitate understanding of the technical scheme of the application, how to construct an asset value prediction model is described below, and how to use the asset value prediction model to predict asset value for a target object is described below.
Method embodiment one
Referring to fig. 1, the figure is a flowchart of a method for constructing an asset value prediction model according to an embodiment of the present application.
The asset value prediction model construction method provided by the embodiment of the application comprises the following steps of S101-S106:
s101: and acquiring training data, actual asset values corresponding to the training data, verification data and actual asset values corresponding to the verification data.
The embodiment of the present application is not limited to the implementation of S101, for example, in one possible implementation, S101 may include: after a large number of training samples are obtained, all the training samples are divided according to a preset division ratio, and training data, verification data and test data are obtained. The preset dividing ratio may be preset, for example, the preset dividing ratio may be 7:2:1.
The training samples refer to data which can be used in the model construction process; and the embodiment of the application does not limit the number of training samples.
In addition, the embodiment of the application also does not limit the acquisition mode of the training sample. For example, if the asset value prediction model is used to predict the asset value of the power distribution network, when determining that the actual asset value corresponding to the training sample is the historical asset value of the power distribution network, the training sample may be determined according to the value impact feature corresponding to the historical asset value of the power distribution network (e.g., the value impact feature corresponding to the historical asset value of the power distribution network is directly determined as the training sample).
The distribution network is a power network which receives electric energy from a power transmission network or a regional power plant and distributes the electric energy to various users locally through a power distribution facility or distributes the electric energy to various users step by step according to voltage.
Value impact characteristics refer to the influencing factors that can affect asset value determinations. In addition, the embodiment of the application does not limit the influence factors of the asset value of the power distribution network, and for example, the influence factors may include at least one of an annual average rate of rentals of the power distribution network service objects, an annual highest air temperature of the power distribution network service objects, an annual lowest air temperature of the power distribution network service objects, an annual average air temperature of the power distribution network service objects, an annual electricity purchase price of the power distribution network, an annual electricity selling price of the power distribution network, a rated power of a power distribution network transformer, a building area of the power distribution network service objects, a number of residents of the power distribution network service objects, a unit area rent of the power distribution network service objects, an average GDP of the power distribution network service objects, a power consumption rate of the power distribution network, and an operation and maintenance cost rate of the power distribution network.
Based on the above-mentioned related content of S101, if the preset division ratio is 7:2:1, after a large number of training samples are obtained, all training samples may be divided into 7/10 training samples, 2/10 training samples and 1/10 training samples by adopting a random sampling manner; and determining the training samples of 7/10 and the corresponding actual asset values as training data and the corresponding actual asset values, and determining the training samples of 2/10 and the corresponding actual asset values as verification data and the corresponding actual asset values.
S102: and inputting the training data and the verification data into a preset neural network model to obtain a first predicted asset value corresponding to the training data and a predicted asset value corresponding to the verification data output by the preset neural network model.
The preset neural network model is used for predicting the asset value, and the embodiment of the application is not limited to the model structure of the preset neural network model. For example, the preset neural network model may include a first hidden layer and a second hidden layer, input data of the second hidden layer includes output data of the first hidden layer, the first hidden layer includes 16 nodes, the second hidden layer includes 8 nodes, and an activation function of each node is a hyperbolic tangent activation function.
In addition, the embodiment of the present application is not limited to the implementation of the node, for example, the node may be implemented by adopting a node structure in a recurrent neural network, where the structure of the node is shown in fig. 2, and the working principle of the node is shown in formulas (1) - (6).
f t ＝sigmoid(W f x t +U f h t-1 +V f C t-1 ) (1)
i t ＝sigmoid(W i x t +U i h t-1 +V i C t-1 ) (2)
G t ＝tanh(W G x t +U G h t-1 ) (3)
o t ＝sigmoid(W o x t +U o h t-1 +V o C t-1 ) (4)
C t ＝f t *C t-1 +i t *G t (5)
h t ＝o t *tanh(C t ) (6)
In the formula, h t The state at time t is represented; h is a t-1 The state at time t-1 is represented; "x" represents matrix element multiplication; w (W) f 、U f 、V f 、W i 、U i 、V i 、W G 、U G 、W o 、U o 、V o Are hidden layer parameters. It should be noted that, each symbol appearing in the formulas (1) - (6) may be explained by using symbols related to nodes in the existing recurrent neural network, which are not described herein for brevity.
The first predicted asset value corresponding to the training data is a predicted asset value obtained and output by predicting the asset value of the training data by a preset neural network model.
The predicted asset value corresponding to the verification data refers to the predicted asset value obtained and output by predicting the asset value of the verification data by a preset neural network model.
Based on the above-mentioned content related to S102, the asset value prediction may be performed on the training data (and/or the verification data) by using the preset neural network model, so as to obtain and output the first predicted asset value corresponding to the training data (and/or the predicted asset value corresponding to the verification data), so that the asset value prediction performance of the preset neural network model may be determined by using the first predicted asset value corresponding to the training data (and/or the predicted asset value corresponding to the verification data).
S103: and determining model performance evaluation characteristics corresponding to the current model updating times according to the first predicted asset value corresponding to the training data, the actual asset value corresponding to the training data, the predicted asset value corresponding to the verification data and the actual asset value corresponding to the verification data.
The current model updating times refer to model updating times of a preset neural network model before a current round of training process; moreover, embodiments of the present application are not limited to the current model update times, which may be, for example, 0, 1, 2, 3, … ….
It should be noted that, if the number of current model updates is 0, it indicates that the preset neural network model has not undergone model update before the current round of training process, thereby indicating that the preset neural network model has not been trained yet.
The model performance evaluation characteristics corresponding to the current model updating times are used for representing the asset value prediction performance of the preset neural network model in the current training process. In addition, embodiments of the present application are not limited to model performance evaluation features, for example, model performance evaluation features may include predicted performance characterization values corresponding to training data and/or predicted performance characterization values corresponding to verification data.
The prediction performance characterization value corresponding to the training data is used for characterizing the prediction performance of the preset neural network model aiming at the training data; the embodiment of the application does not limit the determination mode of the predicted performance characterization value corresponding to the training data. For example, if the number of training data is M 1 When M is to 1 First predicted asset value and M corresponding to the training data 1 An average absolute error (Mean Absolute Error, MAE) between the actual asset values corresponding to the individual training data is determined as a predicted performance characterization value corresponding to the training data.
It should be noted that, MAE refers to an average value of absolute values of errors between predicted values and actual values of a plurality of predicted objects, and MAE may be calculated by using formula (7).
Wherein MAE (b) 1 ，b 2 ，…，b N ) Representing training sample b 1 To training sample b N Corresponding average absolute error;representing training sample b j Corresponding predicted asset value; />Representing training sample b j Corresponding actual asset value; n represents the number of training samples involved in the average absolute error.
The predicted performance characterization value corresponding to the verification data is used for characterizing the predicted performance of the preset neural network model shown by the verification data; the embodiment of the application does not limit the determination mode of the predicted performance characterization value corresponding to the verification data. For example, if the number of verification data is M 2 When M is to 2 Predictive asset value sum M corresponding to individual validation data 2 And determining the average absolute error between the actual asset values corresponding to the verification data as a predicted performance characterization value corresponding to the verification data.
Based on the above-mentioned related content of S103, for the current training process of the preset neural network model, after the first predicted asset value corresponding to the training data and the predicted asset value corresponding to the verification data are obtained, the model performance evaluation feature corresponding to the current model update times may be determined according to the difference between the first predicted asset value corresponding to the training data and the actual asset value corresponding to the training data, and the difference between the predicted asset value corresponding to the verification data and the actual asset value corresponding to the verification data, so that the model performance evaluation feature may accurately represent that the preset neural network model shows the asset value prediction performance in the current training process.
S104: judging whether a preset stopping condition is met, if so, executing S106; if not, S105 is performed.
The preset stopping condition may be preset, and the embodiment of the present application is not limited to the preset stopping condition. For example, the preset stopping condition may be that the difference between the predicted asset value and the actual asset value is lower than a first threshold, the change rate of the predicted asset value is lower than a second threshold, or the number of updates of the preset neural network model reaches a third threshold.
Based on the above-mentioned related content of S104, for the current training process of the preset neural network model, if it is determined that the preset stopping condition is reached, the asset value prediction model may be constructed directly by using the model performance evaluation feature corresponding to the obtained update times of each model; however, upon determining that the preset stopping condition has not been reached, parameters may be performed on the preset neural network model (e.g., W above f 、U f 、V f 、W i 、U i 、V i 、W G 、U G 、W o 、U o 、V o The model internal parameters), and continuously determining the prediction performance of the updated preset neural network model.
It should be noted that, the embodiment of the present application is not limited to the execution sequence of S103 and S104, and S103 and S104 may be executed sequentially, S104 and S103 may be executed sequentially, and S103 and S104 may be executed simultaneously.
S105: and updating the preset neural network model by using the first predicted asset value corresponding to the training data and the actual asset value corresponding to the training data, and returning to execute S102.
The embodiment of the application is not limited to the updating process of the preset neural network model, and can be implemented by adopting any existing or future method capable of updating the preset neural network model.
S106: and constructing an asset value prediction model according to the model performance evaluation characteristics.
The embodiment of the present application is not limited to the implementation of S106, for example, in one possible implementation, S106 may specifically include S1061-S1063:
s1061: and determining the model updating times corresponding to the model overfitting initial characterization points according to the model performance evaluation characteristics.
The model overfitting initial characterization points are used for characterizing the overfitting phenomenon in the training process of the preset neural network model. In addition, the embodiment of the application does not limit the model overfitting initial characterization point, and the model overfitting initial characterization point can refer to a point on the predicted performance change curve corresponding to the training data and the predicted performance change curve corresponding to the verification data.
In practice, the typical performance of the over-fitting phenomenon is that the asset value prediction performance of the preset neural network model for training data is better and better, but the asset value prediction performance of the preset neural network model for verification data is worse and worse. Therefore, when the asset value prediction performance of the preset neural network model for the training data is still in an increasing state, but the asset value prediction performance of the preset neural network model for the verification data is in a decreasing state, the training process of the preset neural network model starts to generate the fitting phenomenon.
Based on this, the embodiment of the application provides an implementation manner for determining the over-fitting initial characterization points of a model, which specifically may include steps 1 to 3:
step 1: and generating a predicted performance change curve corresponding to the training data according to the predicted performance characterization value corresponding to the training data.
The predicted performance change curve corresponding to the training data is used for describing the change trend of the predicted performance representation value corresponding to the training data along with the model updating times of the preset neural network model.
It can be seen that, assuming that the neural network model has undergone T updates when the preset stop condition is reached, the predicted performance characterization value corresponding to the training data at model update number 0 (i.e., the predicted performance characterization value corresponding to the training data at model update number 0), the predicted performance characterization value corresponding to the training data at model update number 1 (i.e., the predicted performance characterization value corresponding to the training data at model update number 1), the predicted performance characterization value corresponding to the training data at model update number … …, and the predicted performance characterization value corresponding to the training data at model update number T (i.e., the predicted performance characterization value corresponding to the training data at model update number T) (i.e., the predicted performance characterization value corresponding to the training data at model update number 2), the predicted performance characterization value corresponding to the training data at model update number 2 (i.e., the predicted performance characterization value corresponding to the training data at model update number 3), the predicted performance characterization value corresponding to the training data at model update number T (i.e., the predicted performance characterization value corresponding to model update number T) can be plotted as the coordinate change of the predicted performance characterization curve (i.e., the predicted performance characterization value corresponding to the training data at model update number 1) or the training number of the training time (i.e., the vertical axis) can be plotted, therefore, the predicted performance change curve corresponding to the training data is used for describing the change trend of the predicted performance representation value corresponding to the training data along with the model updating times of the preset neural network model.
Step 2: and generating a predicted performance change curve corresponding to the verification data according to the predicted performance characterization value corresponding to the verification data.
The predicted performance change curve corresponding to the verification data is used for describing the change trend of the predicted performance characterization value corresponding to the verification data along with the model updating times of the preset neural network model.
It can be seen that, assuming that the neural network model has undergone T updates when the preset stop condition is reached, the predicted performance characterization value corresponding to the verification data at model update number 0 (i.e., the predicted performance characterization value corresponding to the verification data at model update number 0, i.e., the predicted performance characterization value corresponding to the verification data at training process 1), the predicted performance characterization value corresponding to the verification data at model update number 1 (i.e., the predicted performance characterization value corresponding to the verification data at model update number 1, i.e., the predicted performance characterization value corresponding to the verification data at training process 2), the predicted performance characterization value corresponding to the verification data at model update number 2 (i.e., the predicted performance characterization value corresponding to the verification data at model update number 2), … …, and the predicted performance characterization value corresponding to the verification data at model update number T (i.e., the predicted performance characterization value corresponding to the verification data at model update number T), the predicted performance characterization value corresponding to the model update number T (i.e., the predicted performance characterization value corresponding to the verification data at model update number 2), the predicted performance characterization value corresponding to the model update number 2, the predicted performance characterization value corresponding to the model update number T, and the predicted performance characterization value corresponding to the model change (i.e., the model change as the predicted performance characterization value) can be plotted in the coefficient, therefore, the predicted performance change curve corresponding to the verification data is used for describing the change trend of the predicted performance characterization value corresponding to the verification data along with the model updating times of the preset neural network model.
Step 3: and determining a model overfitting initial characterization point meeting a preset overfitting initial condition according to the predicted performance change curve corresponding to the training data and the predicted performance change curve corresponding to the verification data.
The preset overfitting initial condition refers to a condition satisfied by the model overfitting initial characterization point, and the preset overfitting initial condition can be preset. For example, the preset overfitting start condition may refer to that the slope corresponding to the same horizontal axis variable in the predicted performance change curve corresponding to the training data is a positive value, but the slope corresponding to the predicted performance change curve corresponding to the verification data is a negative value.
Therefore, after the predicted performance change curve corresponding to the training data and the predicted performance change curve corresponding to the verification data are obtained, the predicted performance change curve corresponding to the training data and the predicted performance change curve corresponding to the verification data may be compared first, so that the points satisfying the following three conditions are searched in the predicted performance change curve corresponding to the training data and the predicted performance change curve corresponding to the verification data, and the three conditions are: the corresponding slope in the predicted performance change curve corresponding to the training data is a positive value, and the corresponding slope in the predicted performance change curve corresponding to the verification data is a negative value, which have the same abscissa; and then, taking the searched points as model overfitting initial characterization points so as to be convenient for later determining corresponding model updating times when the overfitting phenomenon starts to appear in the training process of the preset neural network model based on the model overfitting initial characterization points.
In addition, the number of model updates corresponding to the model overfitting initial characterization point refers to a point having a correspondence with the model overfitting initial characterization point. For example, if the model overfitting initial characterization point is a point on the predicted performance change curve corresponding to the training data and the predicted performance change curve corresponding to the verification data, the abscissa of the model overfitting initial characterization point is the model update number corresponding to the model overfitting initial characterization point.
Based on the above-mentioned related content of S1062, after it is determined that the preset stopping condition is reached, the model overfitting start characterization point may be determined according to the model performance evaluation feature (for example, the predicted performance characterization value corresponding to the training data and the predicted performance characterization value corresponding to the verification data) having a corresponding relationship with each model update frequency, so that the corresponding model update frequency when the overfitting phenomenon starts to occur in the training process of the preset neural network model can be determined according to the model update frequency corresponding to the model overfitting start characterization point.
S1062: and determining the updating times of the target model according to the model updating times corresponding to the model overfitting initial characterization points.
The target model update times refer to model update times required to be achieved for constructing an asset value prediction model.
In addition, the embodiment of the present application is not limited to the implementation of S1062, and for ease of understanding, the following description will be given with reference to two examples.
Example one, S1062 may specifically be: subtracting 1 from the model updating times corresponding to the model fitting initial characterization points to obtain the target model updating times.
Example two, S1062 may specifically be: and determining the model updating times corresponding to the model overfitting initial characterization points as target model updating times.
Based on the related content of S1062, after the number of model updates corresponding to the model overfitting initial characterization point is obtained, the number of target model updates can be determined according to the number of model updates corresponding to the model overfitting initial characterization point, so that the asset value prediction model can be constructed based on the number of target model updates, and the overfitting phenomenon of the constructed asset value prediction model can be effectively avoided.
S1063: and constructing an asset value prediction model according to the update times of the target model.
The embodiment of the present application is not limited to the implementation of S1063, for example, in one possible implementation, S1063 may specifically include S10631-S10635:
s10631: and determining the untrained preset neural network model as a model to be used.
It should be noted that, the untrained preset neural network model refers to a preset neural network model that has not undergone the above model update process. It can be seen that the untrained pre-set neural network model refers to the pre-set neural network model used in the training process of round 1 above.
S10632: and inputting the training data into the model to be used to obtain a second predicted asset value corresponding to the training data output by the model to be used.
The second predicted asset value corresponding to the training data refers to a predicted asset value obtained and output by predicting the asset value of the training data by using the model.
S10633: judging whether the current updating times of the model to be used reach the target model updating times, if so, executing S10635; if not, then S10634 is performed.
S10634: and updating the model to be used according to the second predicted asset value corresponding to the training data and the actual asset value corresponding to the training data, and returning to execute S1062.
S10635: the model to be used is determined as an asset value prediction model.
Based on the above-mentioned content related to S10631 to S10635, after the number of target model updates is obtained, the training data may be used to update the number of untrained preset neural network models, so that the updated model to be used does not have an overfitting phenomenon, and thus the asset value prediction model determined based on the updated model to be used also does not have an overfitting phenomenon, which is beneficial to improving the asset value prediction accuracy of the asset value prediction model.
Based on the above-mentioned content related to S101 to S106, in the method for constructing an asset value prediction model provided in the embodiment of the present application, after obtaining training data, an actual asset value corresponding to the training data, verification data, and an actual asset value corresponding to the verification data, the training data and the verification data are input into a preset neural network model, so as to obtain a first predicted asset value corresponding to the training data and a predicted asset value corresponding to the verification data output by the preset neural network model; and determining model performance evaluation characteristics corresponding to the current model update times according to the first predicted asset value corresponding to the training data, the actual asset value corresponding to the training data, the predicted asset value corresponding to the verification data and the actual asset value corresponding to the verification data, updating a preset neural network model by using the first predicted asset value corresponding to the training data and the actual asset value corresponding to the training data, continuously executing the input of the training data and the verification data into the preset neural network model, and obtaining the predicted asset value corresponding to the training data and the predicted asset value corresponding to the verification data output by the preset neural network model and subsequent steps until the predicted asset value reaches the preset stop condition is determined, and constructing an asset value prediction model according to the model performance evaluation characteristics.
Therefore, the asset value prediction model is constructed according to the training data, the actual asset value corresponding to the training data, the verification data and the actual asset value corresponding to the verification data, so that the asset value prediction model obtained by construction has good asset value prediction performance, and the asset value prediction process based on the asset value prediction model is more accurate.
After the asset value prediction model is built, asset value prediction may be performed using the asset value prediction model. Based on this, the embodiment of the application further provides an asset value prediction method, and the asset value prediction method is described below with reference to the accompanying drawings.
Method embodiment II
Referring to fig. 3, a flowchart of an asset value prediction method according to an embodiment of the present application is shown.
The asset value prediction method provided by the embodiment of the application comprises the following steps of S301-S302:
s301: and acquiring the value influence characteristics of the target object.
Wherein, the target object refers to an object needing asset value prediction; moreover, embodiments of the present application are not limited to a target object, which may be a power distribution network, for example.
The value influence characteristic of the target object refers to a factor that can influence the determination of the asset value of the target object. For example, the value impact characteristics of the target object may refer to an annual average rate of rentals of the power distribution network service object, an annual highest air temperature of the power distribution network service object, an annual lowest air temperature of the power distribution network service object, an annual average air temperature of the power distribution network service object, an annual power purchase price of the power distribution network, an annual power selling price of the power distribution network, a rated power of a power distribution network transformer, a building area of the power distribution network service object, a number of residents of the power distribution network service object, a unit area rent of the power distribution network service object, a local average GDP of the power distribution network service object, a power consumption rate of the power distribution network, and an operation and maintenance rate of the power distribution network at the last year from the current moment.
S302: and inputting the value influence characteristics of the target object into a pre-constructed asset value prediction model to obtain the predicted asset value of the target object output by the asset value prediction model.
The asset value prediction model can be constructed by utilizing any implementation mode of the asset value prediction model construction method provided by the embodiment of the application.
Based on the above-mentioned related content of S301 to S302, after the value influence feature of the target object is obtained, the value influence feature of the target object may be directly input into the constructed asset value prediction model, so that the asset value prediction model may accurately predict the predicted asset value of the target object according to the value influence feature of the target object, which is beneficial to improving the prediction accuracy of the asset value.
It should be noted that, in some cases, after the predicted asset value of the target object is obtained, the predicted asset value of the target object may also be displayed to the user in a preset manner, so that the user may learn the predicted asset value of the target object.
Based on the asset value prediction model construction method provided by the method embodiment, the embodiment of the application also provides an asset value prediction model construction device, and the asset value prediction model construction device is explained and illustrated below with reference to the accompanying drawings.
Device embodiment 1
For technical details of the asset value prediction model construction device provided by the device embodiment, please refer to the above method embodiment.
Referring to fig. 4, the structure diagram of an asset value prediction model construction device according to an embodiment of the present application is shown.
The asset value prediction model construction device 400 provided by the embodiment of the application comprises:
a first obtaining unit 401, configured to obtain training data, an actual asset value corresponding to the training data, the verification data, and an actual asset value corresponding to the verification data;
a first prediction unit 402, configured to input the training data and the verification data into a preset neural network model, to obtain a first predicted asset value corresponding to the training data and a predicted asset value corresponding to the verification data output by the preset neural network model;
a performance evaluation unit 403, configured to determine a model performance evaluation feature corresponding to the current model update number according to the first predicted asset value corresponding to the training data, the actual asset value corresponding to the training data, the predicted asset value corresponding to the verification data, and the actual asset value corresponding to the verification data;
And a model updating unit 404, configured to update the preset neural network model by using the first predicted asset value corresponding to the training data and the actual asset value corresponding to the training data, and return to the first predicting unit 402 to continue executing the input of the training data and the verification data into the preset neural network model, so as to obtain the first predicted asset value corresponding to the training data and the predicted asset value corresponding to the verification data output by the preset neural network model, until it is determined that a preset stopping condition is reached, and construct an asset value prediction model according to the model performance evaluation feature.
In a possible implementation manner, the model updating unit 404 includes:
the first determining subunit is used for determining the model updating times corresponding to the model overfitting initial characterization points according to the model performance evaluation characteristics; the model overfitting initial characterization points are used for characterizing the overfitting phenomenon in the training process of the preset neural network model;
the second determining subunit is used for determining the target model updating times according to the model updating times corresponding to the model overfitting initial characterization points;
And the model construction subunit is used for constructing an asset value prediction model according to the target model updating times.
In one possible implementation manner, if the model performance evaluation feature includes a predicted performance characterization value corresponding to the training data and a predicted performance characterization value corresponding to the verification data, the determining process of the model over-fitting the initial characterization point includes:
generating a predicted performance change curve corresponding to the training data according to the predicted performance characterization value corresponding to the training data; the predicted performance change curve corresponding to the training data is used for describing the change trend of the predicted performance representation value corresponding to the training data along with the model updating times of the preset neural network model;
generating a predicted performance change curve corresponding to the verification data according to the predicted performance characterization value corresponding to the verification data; the predicted performance change curve corresponding to the verification data is used for describing the change trend of the predicted performance characterization value corresponding to the verification data along with the model updating times of the preset neural network model;
and determining a model overfitting initial characterization point meeting a preset overfitting initial condition according to the predicted performance change curve corresponding to the training data and the predicted performance change curve corresponding to the verification data.
In a possible embodiment, the second determining subunit is specifically configured to:
subtracting 1 from the model updating times corresponding to the model fitting initial characterization points to obtain target model updating times;
or,
and determining the model updating times corresponding to the model fitting initial characterization points as target model updating times.
In a possible embodiment, the model building subunit is specifically configured to:
determining an untrained preset neural network model as a model to be used;
inputting the training data into the model to be used to obtain a second predicted asset value corresponding to the training data output by the model to be used;
updating the model to be used according to the second predicted asset value corresponding to the training data and the actual asset value corresponding to the training data, and continuously executing the step of inputting the training data into the model to be used to obtain the second predicted asset value corresponding to the training data output by the model to be used and subsequent steps until the model to be used is determined to be an asset value prediction model when the current updating times of the model to be used is determined to reach the target model updating times.
In one possible implementation manner, if the model performance evaluation feature includes a predicted performance characterization value corresponding to the training data and a predicted performance characterization value corresponding to the verification data, the determining process of the predicted performance characterization value corresponding to the training data includes:
when the training is performedThe number of training data is M 1 When M is to 1 First predicted asset value corresponding to the training data and the M 1 Determining an average absolute error between actual asset values corresponding to the training data as a predicted performance characterization value corresponding to the training data;
and/or the number of the groups of groups,
the determining process of the predicted performance characterization value corresponding to the verification data comprises the following steps:
when the number of the verification data is M 2 When M is to 2 Predicted asset value corresponding to individual validation data and the M 2 And determining the average absolute error between the actual asset values corresponding to the verification data as a predicted performance characterization value corresponding to the verification data.
In one possible implementation manner, the preset neural network model includes a first hidden layer and a second hidden layer, input data of the second hidden layer includes output data of the first hidden layer, the first hidden layer includes 16 nodes, the second hidden layer includes 8 nodes, and an activation function of each node is a hyperbolic tangent activation function.
Based on the related content of the asset value prediction model construction device 400, after acquiring the training data, the actual asset value corresponding to the training data, the verification data and the actual asset value corresponding to the verification data, inputting the training data and the verification data into a preset neural network model to obtain a first predicted asset value corresponding to the training data and a predicted asset value corresponding to the verification data output by the preset neural network model; and determining model performance evaluation characteristics corresponding to the current model update times according to the first predicted asset value corresponding to the training data, the actual asset value corresponding to the training data, the predicted asset value corresponding to the verification data and the actual asset value corresponding to the verification data, updating a preset neural network model by using the first predicted asset value corresponding to the training data and the actual asset value corresponding to the training data, continuously executing the input of the training data and the verification data into the preset neural network model, and obtaining the predicted asset value corresponding to the training data and the predicted asset value corresponding to the verification data output by the preset neural network model and subsequent steps until the predicted asset value reaches the preset stop condition is determined, and constructing an asset value prediction model according to the model performance evaluation characteristics. Therefore, the asset value prediction model is constructed according to the training data, the actual asset value corresponding to the training data, the verification data and the actual asset value corresponding to the verification data, so that the asset value prediction model obtained by construction has good asset value prediction performance, and the asset value prediction process based on the asset value prediction model is more accurate.
Based on the asset value prediction method provided by the method embodiment, the embodiment of the application also provides an asset value prediction device, which is explained and illustrated below with reference to the accompanying drawings.
Device example two
For technical details of the asset value predicting device provided in the device embodiment, please refer to the above method embodiment.
Referring to fig. 5, the structure of an asset value prediction device according to an embodiment of the present application is shown.
The asset value prediction device 500 provided by the embodiment of the application comprises:
a second acquisition unit 501 for acquiring a value influence feature of the target object;
a second prediction unit 502, configured to input a value influence feature of the target object into a pre-constructed asset value prediction model, and obtain a predicted asset value of the target object output by the asset value prediction model; the asset value prediction model is built by any implementation mode of the asset value prediction model building method provided by the embodiment of the application.
Based on the above-mentioned related content of the asset value predicting device 500, after the value influence feature of the target object is obtained, the value influence feature of the target object may be directly input into the constructed asset value predicting model, so that the asset value predicting model may accurately predict the predicted asset value of the target object according to the value influence feature of the target object, which is beneficial to improving the prediction accuracy of the asset value.
Further, an embodiment of the present application also provides an apparatus, where the apparatus includes a processor and a memory:
the memory is used for storing a computer program;
the processor is configured to execute any implementation mode of the asset value prediction model construction method provided by the embodiment of the present application according to the computer program, or execute any implementation mode of the asset value prediction method provided by the embodiment of the present application.
Further, the embodiment of the application also provides a computer readable storage medium, which is used for storing a computer program, and the computer program is used for executing any implementation mode of the asset value prediction model construction method provided by the embodiment of the application or executing any implementation mode of the asset value prediction method provided by the embodiment of the application.
Further, the embodiment of the application also provides a computer program product, which is characterized in that when the computer program product runs on a terminal device, the terminal device is caused to execute any implementation mode of the asset value prediction model construction method provided by the embodiment of the application or execute any implementation mode of the asset value prediction method provided by the embodiment of the application.
It should be understood that in the present application, "at least one (item)" means one or more, and "a plurality" means two or more. "and/or" for describing the association relationship of the association object, the representation may have three relationships, for example, "a and/or B" may represent: only a, only B and both a and B are present, wherein a, B may be singular or plural. The character "/" generally indicates that the context-dependent object is an "or" relationship. "at least one of" or the like means any combination of these items, including any combination of single item(s) or plural items(s). For example, at least one (one) of a, b or c may represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", wherein a, b, c may be single or plural.
The above description is only of the preferred embodiment of the present application, and is not intended to limit the present application in any way. While the application has been described with reference to preferred embodiments, it is not intended to be limiting. Any person skilled in the art can make many possible variations and modifications to the technical solution of the present application or modifications to equivalent embodiments using the methods and technical contents disclosed above, without departing from the scope of the technical solution of the present application. Therefore, any simple modification, equivalent variation and modification of the above embodiments according to the technical substance of the present application still fall within the scope of the technical solution of the present application.

Claims (11)

1. An asset value prediction model construction method, comprising:
acquiring training data, actual asset values corresponding to the training data, verification data and actual asset values corresponding to the verification data;
inputting the training data and the verification data into a preset neural network model to obtain a first predicted asset value corresponding to the training data and a predicted asset value corresponding to the verification data output by the preset neural network model;
determining model performance evaluation characteristics corresponding to the current model updating times according to the first predicted asset value corresponding to the training data, the actual asset value corresponding to the training data, the predicted asset value corresponding to the verification data and the actual asset value corresponding to the verification data; the model performance evaluation characteristics comprise predicted performance characterization values corresponding to the training data and/or predicted performance characterization values corresponding to the verification data;
updating the preset neural network model by using the first predicted asset value corresponding to the training data and the actual asset value corresponding to the training data, and continuously executing the steps of inputting the training data and the verification data into the preset neural network model to obtain the first predicted asset value corresponding to the training data and the predicted asset value corresponding to the verification data output by the preset neural network model and subsequent steps until the predicted asset value reaches a preset stopping condition, and constructing an asset value prediction model according to the model performance evaluation characteristics;
The construction of the asset value prediction model according to the model performance evaluation characteristics comprises the following steps:
determining the model updating times corresponding to the model overfitting initial characterization points according to the model performance evaluation characteristics; the model overfitting initial characterization points are used for characterizing the overfitting phenomenon in the training process of the preset neural network model;
determining the updating times of the target model according to the model updating times corresponding to the model overfitting initial characterization points;
and constructing an asset value prediction model according to the updating times of the target model.
2. The method of claim 1, wherein if the model performance assessment feature comprises a predicted performance characterization value corresponding to the training data and a predicted performance characterization value corresponding to the verification data, the determining of the model over-fitting a starting characterization point comprises:
generating a predicted performance change curve corresponding to the training data according to the predicted performance characterization value corresponding to the training data; the predicted performance change curve corresponding to the training data is used for describing the change trend of the predicted performance representation value corresponding to the training data along with the model updating times of the preset neural network model;
Generating a predicted performance change curve corresponding to the verification data according to the predicted performance characterization value corresponding to the verification data; the predicted performance change curve corresponding to the verification data is used for describing the change trend of the predicted performance characterization value corresponding to the verification data along with the model updating times of the preset neural network model;
and determining a model overfitting initial characterization point meeting a preset overfitting initial condition according to the predicted performance change curve corresponding to the training data and the predicted performance change curve corresponding to the verification data.
3. The method according to claim 1, wherein determining the number of target model updates according to the number of model updates corresponding to the model overfitting start characterization point comprises:
subtracting 1 from the model updating times corresponding to the model fitting initial characterization points to obtain target model updating times;
or,
and determining the model updating times corresponding to the model fitting initial characterization points as target model updating times.
4. The method of claim 1, wherein constructing an asset value prediction model based on the number of target model updates comprises:
Determining an untrained preset neural network model as a model to be used;
inputting the training data into the model to be used to obtain a second predicted asset value corresponding to the training data output by the model to be used;
updating the model to be used according to the second predicted asset value corresponding to the training data and the actual asset value corresponding to the training data, and continuously executing the step of inputting the training data into the model to be used to obtain the second predicted asset value corresponding to the training data output by the model to be used and subsequent steps until the model to be used is determined to be an asset value prediction model when the current updating times of the model to be used is determined to reach the target model updating times.
5. The method of claim 1, wherein if the model performance assessment feature comprises a predicted performance characterization value corresponding to the training data and a predicted performance characterization value corresponding to the verification data, the determining of the predicted performance characterization value corresponding to the training data comprises:
when the number of the training data is M 1 When M is to 1 First predicted asset value corresponding to the training data and the M 1 Determining an average absolute error between actual asset values corresponding to the training data as a predicted performance characterization value corresponding to the training data;
and/or the number of the groups of groups,
the determining process of the predicted performance characterization value corresponding to the verification data comprises the following steps:
when the number of the verification data is M 2 When M is to 2 Predicted asset value corresponding to individual validation data and the M 2 And determining the average absolute error between the actual asset values corresponding to the verification data as a predicted performance characterization value corresponding to the verification data.
6. The method of claim 1, wherein the pre-set neural network model comprises a first hidden layer and a second hidden layer, wherein the input data of the second hidden layer comprises the output data of the first hidden layer, wherein the first hidden layer comprises 16 nodes, wherein the second hidden layer comprises 8 nodes, and wherein the activation function of each node is a hyperbolic tangent activation function.
7. A method of asset value prediction, the method comprising:
acquiring value influence characteristics of a target object;
inputting the value influence characteristics of the target object into a pre-constructed asset value prediction model to obtain the predicted asset value of the target object output by the asset value prediction model; wherein the asset value prediction model is constructed using the asset value prediction model construction method of any one of claims 1 to 6.
8. An asset value prediction model construction apparatus, comprising:
the first acquisition unit is used for acquiring training data, actual asset values corresponding to the training data, verification data and actual asset values corresponding to the verification data;
the first prediction unit is used for inputting the training data and the verification data into a preset neural network model to obtain a first predicted asset value corresponding to the training data and a predicted asset value corresponding to the verification data output by the preset neural network model;
the performance evaluation unit is used for determining model performance evaluation characteristics corresponding to the current model updating times according to the first predicted asset value corresponding to the training data, the actual asset value corresponding to the training data, the predicted asset value corresponding to the verification data and the actual asset value corresponding to the verification data; the model performance evaluation characteristics comprise predicted performance characterization values corresponding to the training data and/or predicted performance characterization values corresponding to the verification data;
the model updating unit is used for updating the preset neural network model by utilizing the first predicted asset value corresponding to the training data and the actual asset value corresponding to the training data, returning to the first predicting unit to continuously execute the input of the training data and the verification data into the preset neural network model to obtain the first predicted asset value corresponding to the training data and the predicted asset value corresponding to the verification data output by the preset neural network model, and constructing an asset value predicting model according to the model performance evaluation characteristics after the fact that the preset stopping condition is confirmed to be reached;
The model updating unit includes:
the first determining subunit is used for determining the model updating times corresponding to the model overfitting initial characterization points according to the model performance evaluation characteristics; the model overfitting initial characterization points are used for characterizing the overfitting phenomenon in the training process of the preset neural network model;
the second determining subunit is used for determining the target model updating times according to the model updating times corresponding to the model overfitting initial characterization points;
and the model construction subunit is used for constructing an asset value prediction model according to the target model updating times.
9. An asset value prediction apparatus, comprising:
a second acquisition unit configured to acquire a value influence feature of the target object;
the second prediction unit is used for inputting the value influence characteristics of the target object into a pre-constructed asset value prediction model to obtain the predicted asset value of the target object output by the asset value prediction model; wherein the asset value prediction model is constructed using the asset value prediction model construction method of any one of claims 1 to 6.
10. An apparatus comprising a processor and a memory:
The memory is used for storing a computer program;
the processor is configured to execute the asset value prediction model construction method according to any one of claims 1 to 6 or the asset value prediction method according to claim 7 according to the computer program.
11. A computer-readable storage medium storing a computer program for executing the asset value prediction model construction method according to any one of claims 1 to 6 or executing the asset value prediction method according to claim 7.
CN202110206060.9A 2021-02-24 2021-02-24 Asset value prediction model construction method and asset value prediction method Active CN112927013B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110206060.9A CN112927013B (en) 2021-02-24 2021-02-24 Asset value prediction model construction method and asset value prediction method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110206060.9A CN112927013B (en) 2021-02-24 2021-02-24 Asset value prediction model construction method and asset value prediction method

Publications (2)

Publication Number Publication Date
CN112927013A CN112927013A (en) 2021-06-08
CN112927013B true CN112927013B (en) 2023-11-10

Family

ID=76171563

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110206060.9A Active CN112927013B (en) 2021-02-24 2021-02-24 Asset value prediction model construction method and asset value prediction method

Country Status (1)

Country Link
CN (1) CN112927013B (en)

Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108256630A (en) * 2018-01-23 2018-07-06 深圳市唯特视科技有限公司 A kind of over-fitting solution based on low dimensional manifold regularization neural network
CN109840588A (en) * 2019-01-04 2019-06-04 平安科技（深圳）有限公司 Neural network model training method, device, computer equipment and storage medium
CN110414140A (en) * 2019-07-30 2019-11-05 同济大学 The model training and test method of wood damage, device, equipment and storage medium
CN110516804A (en) * 2019-08-22 2019-11-29 广东浪潮大数据研究有限公司 Model training method and device
US10565471B1 (en) * 2019-03-07 2020-02-18 Capital One Services, Llc Systems and methods for transfer learning of neural networks
CN110956251A (en) * 2018-09-26 2020-04-03 北京嘀嘀无限科技发展有限公司 Model training method and device
WO2020097217A1 (en) * 2018-11-06 2020-05-14 Emory University Systems and Methods for Training an Autoencoder Neural Network Using Sparse Data
CN111222677A (en) * 2019-10-22 2020-06-02 浙江运达风电股份有限公司 Wind speed prediction method and system based on long-short term memory time neural network
CN111508000A (en) * 2020-04-14 2020-08-07 北京交通大学 Deep reinforcement learning target tracking method based on parameter space noise network
CN111507499A (en) * 2019-12-05 2020-08-07 阿里巴巴集团控股有限公司 Construction method, test method, device and system of model for prediction
CN111539479A (en) * 2020-04-27 2020-08-14 北京百度网讯科技有限公司 Method and device for generating sample data
CN111798263A (en) * 2020-05-22 2020-10-20 北京国电通网络技术有限公司 Transaction trend prediction method and device
WO2020224297A1 (en) * 2019-05-05 2020-11-12 创新先进技术有限公司 Method and device for determining computer-executable integrated model
CN112037063A (en) * 2020-09-01 2020-12-04 国网雄安金融科技集团有限公司 Exchange rate prediction model generation method, exchange rate prediction method and related equipment
WO2021007812A1 (en) * 2019-07-17 2021-01-21 深圳大学 Deep neural network hyperparameter optimization method, electronic device and storage medium

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030182250A1 (en) * 2002-03-19 2003-09-25 Mohammad Shihidehpour Technique for forecasting market pricing of electricity
US20060247798A1 (en) * 2005-04-28 2006-11-02 Subbu Rajesh V Method and system for performing multi-objective predictive modeling, monitoring, and update for an asset
US20150134401A1 (en) * 2013-11-09 2015-05-14 Carsten Heuer In-memory end-to-end process of predictive analytics
US10769532B2 (en) * 2017-04-05 2020-09-08 Accenture Global Solutions Limited Network rating prediction engine
US11537841B2 (en) * 2019-04-08 2022-12-27 Samsung Electronics Co., Ltd. System and method for compact neural network modeling of transistors

Patent Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108256630A (en) * 2018-01-23 2018-07-06 深圳市唯特视科技有限公司 A kind of over-fitting solution based on low dimensional manifold regularization neural network
CN110956251A (en) * 2018-09-26 2020-04-03 北京嘀嘀无限科技发展有限公司 Model training method and device
WO2020097217A1 (en) * 2018-11-06 2020-05-14 Emory University Systems and Methods for Training an Autoencoder Neural Network Using Sparse Data
CN109840588A (en) * 2019-01-04 2019-06-04 平安科技（深圳）有限公司 Neural network model training method, device, computer equipment and storage medium
US10565471B1 (en) * 2019-03-07 2020-02-18 Capital One Services, Llc Systems and methods for transfer learning of neural networks
WO2020224297A1 (en) * 2019-05-05 2020-11-12 创新先进技术有限公司 Method and device for determining computer-executable integrated model
WO2021007812A1 (en) * 2019-07-17 2021-01-21 深圳大学 Deep neural network hyperparameter optimization method, electronic device and storage medium
CN110414140A (en) * 2019-07-30 2019-11-05 同济大学 The model training and test method of wood damage, device, equipment and storage medium
CN110516804A (en) * 2019-08-22 2019-11-29 广东浪潮大数据研究有限公司 Model training method and device
CN111222677A (en) * 2019-10-22 2020-06-02 浙江运达风电股份有限公司 Wind speed prediction method and system based on long-short term memory time neural network
CN111507499A (en) * 2019-12-05 2020-08-07 阿里巴巴集团控股有限公司 Construction method, test method, device and system of model for prediction
CN111508000A (en) * 2020-04-14 2020-08-07 北京交通大学 Deep reinforcement learning target tracking method based on parameter space noise network
CN111539479A (en) * 2020-04-27 2020-08-14 北京百度网讯科技有限公司 Method and device for generating sample data
CN111798263A (en) * 2020-05-22 2020-10-20 北京国电通网络技术有限公司 Transaction trend prediction method and device
CN112037063A (en) * 2020-09-01 2020-12-04 国网雄安金融科技集团有限公司 Exchange rate prediction model generation method, exchange rate prediction method and related equipment

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
"基于神经网络模型的人民币汇率预测研究";丁晖;《中国优秀博士学位论文全文数据库》;1-190 *

Also Published As

Publication number Publication date
CN112927013A (en) 2021-06-08

Similar Documents

Publication Publication Date Title
CN105391083B (en) Wind power interval short term prediction method based on variation mode decomposition and Method Using Relevance Vector Machine
CN106533750B (en) The forecasting system and method for non-stationary type application user concurrent amount under a kind of cloud environment
CN110781947B (en) Power load prediction model training and power load prediction method and device
CN112381673B (en) Park electricity utilization information analysis method and device based on digital twin
CN109379240A (en) Car networking flux prediction model construction method, device and electronic equipment
CN112418921A (en) Power demand prediction method, device, system and computer storage medium
CN110781969A (en) Air conditioner air volume control method and device based on deep reinforcement learning and medium
CN111523084A (en) Service data prediction method and device, electronic equipment and computer readable storage medium
CN110766236A (en) Power equipment state trend prediction method based on statistical analysis and deep learning
CN116307215A (en) Load prediction method, device, equipment and storage medium of power system
CN112365056A (en) Electrical load joint prediction method and device, terminal and storage medium
CN116227637A (en) Active power distribution network oriented refined load prediction method and system
CN112580896A (en) Knowledge point prediction method, knowledge point prediction device, knowledge point prediction equipment and storage medium
CN106295877B (en) Method for predicting electric energy consumption of smart power grid
CN110967471A (en) Method for predicting concentration of dissolved gas in transformer oil
CN113902260A (en) Information prediction method, information prediction device, electronic equipment and medium
CN112927013B (en) Asset value prediction model construction method and asset value prediction method
CN113393034A (en) Electric quantity prediction method of online self-adaptive OSELM-GARCH model
CN115622034A (en) Internet of things meter power load identification method based on improved genetic neural network
CN112561575B (en) CTR (China railway) prediction model selection method and device
CN114169763A (en) Measuring instrument demand prediction method, system, computing device and storage medium
CN114997510A (en) PM2.5 prediction method and system based on Copula entropy correlation analysis
CN114545270A (en) Lithium battery health state estimation method considering health factor delay
CN113537693A (en) Personnel risk level obtaining method, terminal and storage device
CN112417767A (en) Attenuation trend determination model construction method and attenuation trend determination method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
CB02 Change of applicant information
CB02 Change of applicant information

Address after: 100032 room 8018, 8 / F, building 7, Guangyi street, Xicheng District, Beijing

Applicant after: State Grid Digital Technology Holdings Co.,Ltd.

Applicant after: Guowang Xiongan Finance Technology Group Co.,Ltd.

Address before: 100038 room 8018, 8 / F, building 7, Guangyi street, Xicheng District, Beijing

Applicant before: STATE GRID ELECTRONIC COMMERCE Co.,Ltd.

Applicant before: Guowang Xiongan Finance Technology Group Co.,Ltd.

GR01 Patent grant
GR01 Patent grant