JPWO2022134480A5 - - Google Patents
Download PDFInfo
- Publication number
- JPWO2022134480A5 JPWO2022134480A5 JP2021576944A JP2021576944A JPWO2022134480A5 JP WO2022134480 A5 JPWO2022134480 A5 JP WO2022134480A5 JP 2021576944 A JP2021576944 A JP 2021576944A JP 2021576944 A JP2021576944 A JP 2021576944A JP WO2022134480 A5 JPWO2022134480 A5 JP WO2022134480A5
- Authority
- JP
- Japan
- Prior art keywords
- region
- network
- sample
- risk prediction
- risk
- 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.)
- Pending
Links
Description
さらに、トレーニングデータ内の各セルに対して領域特徴をそれぞれ抽出することができる。本開示で抽出された領域特徴は、周辺の予め設定されたタイプのPOI特徴、人口統計学特徴、及びユーザ外出特徴のうちの少なくとも1つを含むことができる。従来の伝染病モデルとは異なり、本開示で使用されるこれらの領域特徴は、確認された症例と関係ないため、前期経験のない疫病未爆発都市は、同様にセルリスクの予測を行うことができる。以下はこれらの特徴についてそれぞれ詳細に説明する。
Furthermore, region features can be extracted for each cell in the training data, respectively. The region features extracted in the present disclosure may include at least one of surrounding preset types of POI features, demographic features, and user out-of-office features. Unlike traditional epidemic models, these regional features used in this disclosure are not related to confirmed cases, so epidemic-unexploded cities with no pre-experience can similarly make predictions of cell risk. can. Each of these features is described in detail below.
ユーザ外出特徴:
いくつかの関連研究によると、ユーザ外出行為は、通常、疫病伝播に密接な関係があることが証明されている。本開示に係るユーザ外出特徴は、以下のうちの少なくとも1つを含むことができるが、限定せず、
第1:外出方式。例えば、歩行、自転車、公共交通、自家用車など外出方式を予め定義することができる。
User Outgoing Features:
Several relevant studies have demonstrated that user outings are usually closely related to epidemic transmission. User exit features according to the present disclosure may include, but are not limited to, at least one of the following:
1st: Go-out method. For example, it is possible to define in advance how to go out, such as walking, cycling, public transportation, and private car.
第3:出発地-外出方式-目的地モード分布。その中、出発地は本セルを指し、外出方式と目的地タイプは、予め定義することができ、次に、本セルの外出方式と目的地タイプで構成される組み合わせのうち、上位N個の組み合わせを特徴として統計する。Nは予め設定された正の整数であり、例えば、20を選択する。
3rd: Origin- outing method-destination mode distribution. Among them, the departure point refers to this cell, and the travel method and destination type can be defined in advance. Statisticalize the combination as a feature. N is a preset positive integer, and selects 20, for example.
サンプルセルから抽出された領域特徴を符号化ネットワークの入力とし、実際のトレーニングプロセスでは異なるリスクレベル都市に属するサンプルセルを使用するため、本実施例では、高リスク都市のサンプルセルと低リスク都市のサンプルセルを例とする。nE
r、nE
h、及びnE
tは、高リスク都市のサンプルセルの周辺の予め設定されたタイプのPOI特徴、人口統計学特徴、及びユーザ外出特徴をそれぞれ示し、nL
r、nL
h、及びnL
tは、低リスク都市のサンプルセルの周辺の予め設定されたタイプのPOI特徴、人口統計学特徴、及びユーザ外出特徴をそれぞれ示す。nE
r、nE
h、及びnE
tに対してスプライシングなどの融合方式を行った後、高リスク都市のサンプルセルの特徴nEを取得する。nL
r、nL
h、及びnL
tに対してスプライシングなどの融合方式を行った後、低リスク都市のサンプルセルの特徴nLを取得する。
The region features extracted from the sample cells are taken as the input of the encoding network, and the actual training process uses sample cells belonging to different risk level cities. Take the sample cell as an example. Let nEr , nEh , and nEt denote preset types of POI features, demographic features , and user out-of-office features, respectively, around the high-risk city sample cell, and nLr , n L h and n L t denote preset types of POI features, demographic features, and user out-of- office features, respectively, around the low-risk city sample cell. After performing a fusion scheme such as splicing on nE r , nEh , and nEt , we obtain the feature nE of the high-risk city sample cell. After performing a fusion scheme such as splicing on nLr , nLh , and nLt , we obtain nL features of sample cells in low-risk cities .
判別モデルの機能は、入力の
The feature of the discriminant model is that of the input
さらに、学習都市間の共通性特徴以外に、判別モデルは、依然として、自体の機能、すなわち出所する都市リスクレベルを認識することを確保する必要がある。従って、対抗学習の方式を使用して、第1の損失関数L1と呼ばれる1つの損失関数を再構築することができ、当該損失関数は、判別モデルをトレーニングして、判別ネットワークがサンプル領域に対する認識結果とラベリング結果との差異を最小化するために用いられる。例えば、以下の内容を使用することができ、
Furthermore, besides the commonality features between learning cities, the discriminant model still needs to ensure that it recognizes its own function, namely the city risk level of origin. Therefore, we can use a method of counter-learning to reconstruct one loss function, called the first loss function L1 , which trains the discriminant model so that the discriminant network is It is used to minimize the difference between recognition results and labeling results. For example, you can use
分類ネットワークは
The classification network is
符号化ネットワークの役割は、入力したサンプルセルの特徴表現を使用して領域特徴を再構築する。すなわちnEを再構築してベクトル表示
The role of the encoding network is to reconstruct the region features using the input sample cell feature representations. i.e., reconstruct nE to represent the vector
本ステップにおいて領域特徴を抽出する方式は、トレーニングリスク予測モデルプロセスで使用される領域特徴と一致する。同様に周辺の予め設定されたタイプのPOI特徴、人口統計学特徴、及びユーザ外出特徴のうちの少なくとも1つを含むことができる。具体的な領域特徴の内容は、図1に示す実施例の関連説明を参照し、ここで詳細に説明しない。
The manner in which the region features are extracted in this step is consistent with the region features used in the training risk prediction model process. Similarly, surrounding preset types of POI characteristics, demographic characteristics, and/or user out-of-office characteristics may be included. The specific content of the area feature is referred to the related description of the embodiment shown in FIG. 1 and will not be described in detail here.
図3に示すように、nT
r、nT
h、及びnT
tを高リスク都市のサンプルセルの周辺の予め設定されたタイプのPOI特徴、人口統計学特徴、及びユーザ外出特徴としてそれぞれ示す。nT
r、nT
h、及びnT
tに対してスプライシングなどの融合方式を行った後、ターゲットセルの特徴nTを取得する。
As shown in FIG. 3, let nT r , nT h , and nT t be the preset types of POI features, demographic features, and user out-of-office features around the high-risk city sample cell, respectively. . After performing a fusion scheme such as splicing on nT r , nT h and nT t , we obtain the feature nT of the target cell.
特徴抽出ユニット503は、周辺の予め設定されたタイプのPOI特徴、人口統計学特徴、及びユーザ外出特徴のうちの少なくとも1つを含むサンプル領域の領域特徴を取得するために用いられる。
The feature extraction unit 503 is used to obtain area features of the sample area including at least one of surrounding preset types of POI features, demographic features, and user out-of-office features.
ユーザ外出特徴は、外出方式、出発地-目的地モード分布、出発地-外出方式-目的地モード分布のうちの少なくとも1つを含む。
The user outing feature includes at least one of out-going manner, origin-destination mode distribution, origin- out-going manner-destination mode distribution.
Claims (20)
サンプル領域セットと、前記サンプル領域セットにおける各サンプル領域のリスクレベルと各サンプル領域が属する地域のリスクレベルのラベリング結果を含むトレーニングデータを取得するステップと、
前記トレーニングデータを使用して符号化ネットワーク、判別ネットワーク、及び分類ネットワークを含む初期モデルをトレーニングし、トレーニングの完了後に前記初期モデル内の符号化ネットワークと分類ネットワークを使用して前記リスク予測モデルを取得するステップと、を含み、
前記符号化ネットワークは、サンプル領域の領域特徴を使用して、符号化して各サンプル領域の特徴表現を取得し、前記判別ネットワークは、サンプル領域の特徴表現に基づいてサンプル領域が属する地域のリスクレベルを認識し、前記分類ネットワークは、サンプル領域の特徴表現に基づいてサンプル領域のリスクレベルを認識し、前記初期モデルのトレーニングターゲットは、前記判別ネットワークが異なるリスクレベル地域に属するサンプル領域に対する認識差異を最小化し、前記分類ネットワークがサンプル領域に対する認識結果とラベリング結果との差異を最小化することを含む、
リスク予測モデルの確立方法。 A method for establishing a risk prediction model, comprising:
obtaining training data including a sample area set and the labeling results of the risk level of each sample area in the sample area set and the risk level of the region to which each sample area belongs;
training an initial model including an encoding network, a discriminant network, and a classification network using the training data, and obtaining the risk prediction model using the encoding network and the classification network in the initial model after training is complete; and
The encoding network uses the region features of the sample regions to encode to obtain a feature representation of each sample region, and the discriminant network determines the risk level of the region to which the sample region belongs based on the feature representation of the sample regions. wherein the classification network recognizes risk levels of sample regions based on sample region feature representations, and the training target of the initial model is the recognition difference for sample regions belonging to different risk level regions to which the discriminant network recognizes wherein the classification network minimizes the difference between recognition and labeling results for sample regions;
How to establish a risk prediction model.
周辺の予め設定されたタイプのPOI特徴、人口統計学特徴、及びユーザ外出特徴のうちの少なくとも1つを含む、
請求項1に記載のリスク予測モデルの確立方法。 The area features of the sample area are:
including at least one of a surrounding preset type of POI characteristics, demographic characteristics, and user out-of-office characteristics;
A method for establishing a risk prediction model according to claim 1.
前記人口統計学特徴は、人口密度状況、通勤距離分布、年齢分布、性別分布、収入分布、消費能力分布、教育レベル分布、結婚状況分布、生活段階分布、就業タイプ分布、及び業界タイプ分布のうちの少なくとも1つを含み、
前記ユーザ外出特徴は、外出方式、出発地-目的地モード分布、出発地-外出方式-目的地モード分布のうちの少なくとも1つを含む、
請求項2に記載のリスク予測モデルの確立方法。 The peripheral preset type POI features are distance information between the sample area and the nearest preset type POI, and the degree of completeness of living facilities within the preset distance range of the sample area. including at least one
The demographic characteristics are population density status, commuting distance distribution, age distribution, gender distribution, income distribution, consumption ability distribution, education level distribution, marital status distribution, life stage distribution, employment type distribution, and industry type distribution. including at least one of
the user outing feature includes at least one of an outing manner, origin-destination mode distribution, origin- outing manner-destination mode distribution;
A method for establishing a risk prediction model according to claim 2.
前記復号ネットワークは、サンプル領域の特徴表現に基づいて領域特徴を再構築し、
前記トレーニングターゲットは、前記復号ネットワークによって再構築された領域特徴とサンプル領域から抽出された領域特徴との差異を最小化することをさらに含む、
請求項1に記載のリスク予測モデルの確立方法。 the initial model further comprising a decoding network;
the decoding network reconstructs region features based on feature representations of sample regions;
The training target further comprises minimizing the difference between the region features reconstructed by the decoding network and the region features extracted from the sample region.
A method for establishing a risk prediction model according to claim 1.
前記第1の損失関数は、前記判別ネットワークがサンプル領域に対する認識結果とラベリング結果との差異を最小化するために用いられ、
前記第2の損失関数は、前記判別ネットワークが異なるリスクレベル地域に属するサンプル領域に対する認識差異を最小化するために用いられ、
前記第3の損失関数は、前記分類ネットワークがサンプル領域に対する認識結果とラベリング結果との差異を最小化するために用いられ、
前記第4の損失関数は、前記復号ネットワークによって再構築された領域特徴とサンプル領域から抽出された領域特徴との差異を最小化するために用いられる、
請求項4に記載のリスク予測モデルの確立方法。 In the process of training the initial model, optimizing parameters of the discriminant network using a first loss function, and using a second loss function, a third loss function, and a fourth loss function to optimizing the parameters of the encoding network, optimizing the parameters of the classification network using the third loss function, and optimizing the parameters of the decoding network using the fourth loss function;
the first loss function is used by the discriminant network to minimize the difference between recognition results and labeling results for sample regions;
The second loss function is used to minimize recognition differences for sample regions belonging to different risk level regions of the discriminant network,
the third loss function is used by the classification network to minimize the difference between recognition and labeling results for sample regions;
the fourth loss function is used to minimize the difference between regional features reconstructed by the decoding network and regional features extracted from sample regions;
A method for establishing a risk prediction model according to claim 4.
ターゲット領域の領域特徴を抽出するステップと、
前記領域特徴をリスク予測モデルに入力し、前記リスク予測モデルによって出力された結果に基づいて前記ターゲット領域のリスクレベルを決定するステップと、を含み、
前記リスク予測モデルは、請求項1~5のいずれか一つに記載のリスク予測モデルの確立方法を使用して予め確立される、
領域リスク予測方法。 A region risk prediction method comprising:
extracting region features of the target region;
inputting the region features into a risk prediction model and determining a risk level for the target region based on results output by the risk prediction model;
The risk prediction model is pre-established using the method for establishing a risk prediction model according to any one of claims 1 to 5,
Region Risk Prediction Method.
請求項6に記載の領域リスク予測方法。 said risk level is a risk level of epidemic transmission;
The region risk prediction method according to claim 6.
サンプル領域セットと、前記サンプル領域セットにおける各サンプル領域のリスクレベルと各サンプル領域が属する地域のリスクレベルのラベリング結果を含むトレーニングデータを取得するためのデータ取得ユニットと、
前記トレーニングデータを使用して符号化ネットワーク、判別ネットワーク、及び分類ネットワークを含む初期モデルをトレーニングし、トレーニングの完了後に前記初期モデル内の符号化ネットワークと分類ネットワークを使用して前記リスク予測モデルを取得するためのモデルトレーニングユニットと、を含み、
前記符号化ネットワークは、サンプル領域の領域特徴を使用して、符号化して各サンプル領域の特徴表現を取得し、前記判別ネットワークは、サンプル領域の特徴表現に基づいてサンプル領域が属する地域のリスクレベルを認識し、前記分類ネットワークは、サンプル領域の特徴表現に基づいてサンプル領域のリスクレベルを認識し、前記初期モデルのトレーニングターゲットは、前記判別ネットワークが異なるリスクレベル地域に属するサンプル領域に対する認識差異を最小化し、前記分類ネットワークがサンプル領域に対する認識結果とラベリング結果との差異を最小化することを含む、
リスク予測モデルの確立装置。 A risk prediction model establishment device comprising:
a data acquisition unit for acquiring training data including a sample area set, a labeling result of the risk level of each sample area in the sample area set and the risk level of the region to which each sample area belongs;
training an initial model including an encoding network, a discriminant network, and a classification network using the training data, and obtaining the risk prediction model using the encoding network and the classification network in the initial model after training is complete; a model training unit for
The encoding network uses the region features of the sample regions to encode to obtain a feature representation of each sample region, and the discriminant network determines the risk level of the region to which the sample region belongs based on the feature representation of the sample regions. wherein the classification network recognizes risk levels of sample regions based on sample region feature representations, and the training target of the initial model is the recognition difference for sample regions belonging to different risk level regions to which the discriminant network recognizes wherein the classification network minimizes the difference between recognition and labeling results for sample regions;
Equipment for establishing risk prediction models.
請求項8に記載のリスク予測モデルの確立装置。 further comprising a feature extraction unit used to obtain area features of the sample area, including at least one of surrounding preset types of POI features, demographic features, and user exit features;
An apparatus for establishing a risk prediction model according to claim 8.
前記人口統計学特徴は、人口密度状況、通勤距離分布、年齢分布、性別分布、収入分布、消費能力分布、教育レベル分布、結婚状況分布、生活段階分布、就業タイプ分布、及び業界タイプ分布のうちの少なくとも1つを含み、
前記ユーザ外出特徴は、外出方式、出発地-目的地モード分布、出発地-外出方式-目的地モード分布のうちの少なくとも1つを含む、
請求項9に記載のリスク予測モデルの確立装置。 The peripheral preset type POI features are distance information between the sample area and the nearest preset type POI, and the degree of completeness of living facilities within the preset distance range of the sample area. including at least one
The demographic characteristics are population density status, commuting distance distribution, age distribution, gender distribution, income distribution, consumption ability distribution, education level distribution, marital status distribution, life stage distribution, employment type distribution, and industry type distribution. including at least one of
the user outing feature includes at least one of an outing manner, origin-destination mode distribution, origin- outing manner-destination mode distribution;
An apparatus for establishing a risk prediction model according to claim 9.
前記復号ネットワークは、サンプル領域の特徴表現に基づいて領域特徴を再構築し、
前記トレーニングターゲットは、前記復号ネットワークによって再構築された領域特徴とサンプル領域から抽出された領域特徴との差異を最小化することをさらに含む、
請求項8に記載のリスク予測モデルの確立装置。 the initial model further comprising a decoding network;
the decoding network reconstructs region features based on feature representations of sample regions;
The training target further comprises minimizing the difference between the region features reconstructed by the decoding network and the region features extracted from the sample region.
An apparatus for establishing a risk prediction model according to claim 8.
前記第1の損失関数は、前記判別ネットワークがサンプル領域に対する認識結果とラベリング結果との差異を最小化するために用いられ、
前記第2の損失関数は、前記判別ネットワークが異なるリスクレベル地域に属するサンプル領域に対する認識差異を最小化するために用いられ、
前記第3の損失関数は、前記分類ネットワークがサンプル領域に対する認識結果とラベリング結果との差異を最小化するために用いられ、
前記第4の損失関数は、前記復号ネットワークによって再構築された領域特徴とサンプル領域から抽出された領域特徴との差異を最小化するために用いられる、
請求項11に記載のリスク予測モデルの確立装置。 In the process of the model training unit training the initial model, optimizing the parameters of the discriminant network using a first loss function, a second loss function, a third loss function, and a fourth loss function. is used to optimize the parameters of the encoding network, the third loss function is used to optimize the parameters of the classification network, and the fourth loss function is used to optimize the parameters of the decoding network. become
the first loss function is used by the discriminant network to minimize the difference between recognition results and labeling results for sample regions;
The second loss function is used to minimize recognition differences for sample regions belonging to different risk level regions of the discriminant network,
the third loss function is used by the classification network to minimize the difference between recognition and labeling results for sample regions;
the fourth loss function is used to minimize the difference between regional features reconstructed by the decoding network and regional features extracted from sample regions;
An apparatus for establishing a risk prediction model according to claim 11.
ターゲット領域の領域特徴を抽出するための特徴抽出ユニットと、
前記領域特徴をリスク予測モデルに入力し、前記リスク予測モデルによって出力された結果に基づいて前記ターゲット領域のリスクレベルを決定するためのリスク予測ユニットと、を含み、
前記リスク予測モデルは、請求項8~12のいずれか一つに記載のリスク予測モデルの確立装置によって予め確立される、
領域リスク予測装置。 A region risk prediction device,
a feature extraction unit for extracting region features of the target region;
a risk prediction unit for inputting the region features into a risk prediction model and determining a risk level of the target region based on results output by the risk prediction model;
The risk prediction model is pre-established by the risk prediction model establishment device according to any one of claims 8 to 12,
Area risk predictor.
請求項13に記載の領域リスク予測装置。 said risk level is a risk level of epidemic transmission;
The region risk prediction device according to claim 13.
少なくとも一つのプロセッサと、
前記少なくとも一つのプロセッサに通信接続されたメモリと、を含み、
前記メモリに前記少なくとも一つのプロセッサにより実行可能な命令が記憶されており、前記命令が前記少なくとも一つのプロセッサにより実行されると、前記少なくとも一つのプロセッサが請求項1~5のいずれか一つに記載のリスク予測モデルの確立方法を実行する、
電子機器。 an electronic device,
at least one processor;
a memory communicatively coupled to the at least one processor;
Instructions executable by the at least one processor are stored in the memory, and when the instructions are executed by the at least one processor, the at least one processor performs any one of claims 1 to 5. Carrying out the described method of establishing a risk prediction model,
Electronics.
少なくとも一つのプロセッサと、
前記少なくとも一つのプロセッサに通信接続されたメモリと、を含み、
前記メモリに前記少なくとも一つのプロセッサにより実行可能な命令が記憶されており、前記命令が前記少なくとも一つのプロセッサにより実行されると、前記少なくとも一つのプロセッサが請求項6又は7に記載の領域リスク予測方法を実行する、
電子機器。 an electronic device,
at least one processor;
a memory communicatively coupled to the at least one processor;
Instructions executable by the at least one processor are stored in the memory, and when the instructions are executed by the at least one processor, the at least one processor performs the area risk prediction according to claim 6 or 7. carry out the method,
Electronics.
前記コンピュータ命令は、コンピュータに請求項1~5のいずれか一つに記載のリスク予測モデルの確立方法を実行させる、
コンピュータ命令が記憶されている非一時的なコンピュータ読み取り可能な記憶媒体。 A non-transitory computer-readable storage medium having computer instructions stored thereon,
The computer instructions cause a computer to execute the method for establishing a risk prediction model according to any one of claims 1 to 5,
A non-transitory computer-readable storage medium on which computer instructions are stored.
前記コンピュータ命令は、コンピュータに請求項6又は7に記載の領域リスク予測方法を実行させる、
コンピュータ命令が記憶されている非一時的なコンピュータ読み取り可能な記憶媒体。 A non-transitory computer-readable storage medium having computer instructions stored thereon,
The computer instructions cause a computer to perform the area risk prediction method of claim 6 or 7,
A non-transitory computer-readable storage medium on which computer instructions are stored.
コンピュータプログラム。 realizing the method for establishing a risk prediction model according to any one of claims 1 to 5 when executed by a processor;
computer program.
コンピュータプログラム。 realizing a region risk prediction method according to claim 6 or 7 when executed by a processor;
computer program.
Applications Claiming Priority (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011515953.3 | 2020-12-21 | ||
CN202011515953.3A CN112508300B (en) | 2020-12-21 | 2020-12-21 | Method for establishing risk prediction model, regional risk prediction method and corresponding device |
PCT/CN2021/097958 WO2022134480A1 (en) | 2020-12-21 | 2021-06-02 | Method for establishing risk prediction model, regional risk prediction method and corresponding apparatus |
Publications (2)
Publication Number | Publication Date |
---|---|
JP2023510665A JP2023510665A (en) | 2023-03-15 |
JPWO2022134480A5 true JPWO2022134480A5 (en) | 2023-07-11 |
Family
ID=74921829
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
JP2021576944A Pending JP2023510665A (en) | 2020-12-21 | 2021-06-02 | Risk prediction model establishment method, area risk prediction method, and response device |
Country Status (6)
Country | Link |
---|---|
US (1) | US20220398465A1 (en) |
EP (1) | EP4040353B1 (en) |
JP (1) | JP2023510665A (en) |
KR (1) | KR20220093046A (en) |
CN (1) | CN112508300B (en) |
WO (1) | WO2022134480A1 (en) |
Families Citing this family (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112508300B (en) * | 2020-12-21 | 2023-04-18 | 北京百度网讯科技有限公司 | Method for establishing risk prediction model, regional risk prediction method and corresponding device |
CN113744888B (en) * | 2021-09-02 | 2023-09-22 | 深圳万海思数字医疗有限公司 | Regional epidemic trend prediction and early warning method and system |
CN113837588B (en) * | 2021-09-17 | 2023-12-29 | 北京百度网讯科技有限公司 | Training method and device for evaluation model, electronic equipment and storage medium |
CN114372642B (en) * | 2022-03-21 | 2022-05-20 | 创意信息技术股份有限公司 | Method for risk assessment of tourist attraction in urban festivals and holidays |
CN115935265B (en) * | 2023-03-03 | 2023-05-26 | 支付宝(杭州)信息技术有限公司 | Method for training risk identification model, risk identification method and corresponding device |
CN115983142B (en) * | 2023-03-21 | 2023-08-29 | 之江实验室 | Regional population evolution model construction method based on depth generation countermeasure network |
CN116028964B (en) * | 2023-03-28 | 2023-05-23 | 中国标准化研究院 | Information security risk management system |
CN117421244B (en) * | 2023-11-17 | 2024-05-24 | 北京邮电大学 | Multi-source cross-project software defect prediction method, device and storage medium |
CN117932487B (en) * | 2023-12-28 | 2024-08-20 | 中信建投证券股份有限公司 | Risk classification model training and risk classification method and device |
Family Cites Families (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10535424B2 (en) * | 2016-02-19 | 2020-01-14 | International Business Machines Corporation | Method for proactive comprehensive geriatric risk screening |
US11468262B2 (en) * | 2017-10-30 | 2022-10-11 | Nec Corporation | Deep network embedding with adversarial regularization |
CN109902880A (en) * | 2019-03-13 | 2019-06-18 | 南京航空航天大学 | A kind of city stream of people's prediction technique generating confrontation network based on Seq2Seq |
CN110458572B (en) * | 2019-07-08 | 2023-11-24 | 创新先进技术有限公司 | User risk determining method and target risk recognition model establishing method |
CN110674979B (en) * | 2019-09-11 | 2024-09-03 | 腾讯科技(深圳)有限公司 | Training method and device of risk prediction model, prediction method and device, medium and equipment |
CN110689184A (en) * | 2019-09-21 | 2020-01-14 | 广东毓秀科技有限公司 | Method for predicting rail traffic stream of people through deep learning |
CN110993119B (en) * | 2020-03-04 | 2020-07-07 | 同盾控股有限公司 | Epidemic situation prediction method and device based on population migration, electronic equipment and medium |
CN111128399B (en) * | 2020-03-30 | 2020-07-14 | 广州地理研究所 | Epidemic disease epidemic situation risk level assessment method based on people stream density |
CN111523596B (en) * | 2020-04-23 | 2023-07-04 | 北京百度网讯科技有限公司 | Target recognition model training method, device, equipment and storage medium |
CN111626119B (en) * | 2020-04-23 | 2023-09-01 | 北京百度网讯科技有限公司 | Target recognition model training method, device, equipment and storage medium |
CN111523597B (en) * | 2020-04-23 | 2023-08-25 | 北京百度网讯科技有限公司 | Target recognition model training method, device, equipment and storage medium |
CN111626490A (en) * | 2020-05-20 | 2020-09-04 | 南京航空航天大学 | Multitask city space-time prediction method based on counterstudy |
CN111768873A (en) * | 2020-06-03 | 2020-10-13 | 中国地质大学(武汉) | COVID-19 real-time risk prediction method |
CN112508300B (en) * | 2020-12-21 | 2023-04-18 | 北京百度网讯科技有限公司 | Method for establishing risk prediction model, regional risk prediction method and corresponding device |
-
2020
- 2020-12-21 CN CN202011515953.3A patent/CN112508300B/en active Active
-
2021
- 2021-06-02 EP EP21820072.3A patent/EP4040353B1/en active Active
- 2021-06-02 US US17/620,820 patent/US20220398465A1/en active Pending
- 2021-06-02 KR KR1020217042672A patent/KR20220093046A/en not_active Application Discontinuation
- 2021-06-02 JP JP2021576944A patent/JP2023510665A/en active Pending
- 2021-06-02 WO PCT/CN2021/097958 patent/WO2022134480A1/en unknown
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6919021B2 (en) | Generative Adversarial Network (GAN) training methods, image generation methods using GAN, and computer-readable storage media | |
CN107861938B (en) | POI (Point of interest) file generation method and device and electronic equipment | |
KR102234850B1 (en) | Method and apparatus for complementing knowledge based on relation network | |
CN114898121B (en) | Automatic generation method for concrete dam defect image description based on graph attention network | |
CN108375379A (en) | The fast path planing method and mobile robot of dual DQN based on variation | |
CN111754532B (en) | Image segmentation model searching method, device, computer equipment and storage medium | |
CN113807222B (en) | Video question-answering method and system for end-to-end training based on sparse sampling | |
CN114820871B (en) | Font generation method, model training method, device, equipment and medium | |
CN114693624B (en) | Image detection method, device, equipment and readable storage medium | |
CN114783608A (en) | Construction method of slow patient group disease risk prediction model based on graph self-encoder | |
CN108959512B (en) | Image description network and technology based on attribute enhanced attention model | |
CN116541779B (en) | Individualized public safety emergency detection model training method, detection method and device | |
CN116363452B (en) | Task model training method and device | |
JPWO2022134480A5 (en) | ||
CN111708871A (en) | Dialog state tracking method and device and dialog state tracking model training method | |
CN114743037A (en) | Deep medical image clustering method based on multi-scale structure learning | |
CN115098692A (en) | Cross-domain recommendation method and device, electronic equipment and storage medium | |
CN108234195B (en) | Method, apparatus, device, medium for predicting network performance | |
CN114626529A (en) | Natural language reasoning fine-tuning method, system, device and storage medium | |
CN116778011B (en) | Image generating method | |
CN116796821A (en) | Efficient neural network architecture searching method and device for 3D target detection algorithm | |
CN108665455B (en) | Method and device for evaluating image significance prediction result | |
CN116978464A (en) | Data processing method, device, equipment and medium | |
CN110728217A (en) | SAR image recognition method, device, equipment and storage medium | |
CN118279113B (en) | Digital intelligence social public service management method and system based on large model |